Black Lives Matter. Support the Equal Justice Initiative.

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  // Equal reports whether a and b
    16  // are the same length and contain the same bytes.
    17  // A nil argument is equivalent to an empty slice.
    18  func Equal(a, b []byte) bool {
    19  	// Neither cmd/compile nor gccgo allocates for these string conversions.
    20  	return string(a) == string(b)
    21  }
    22  
    23  // Compare returns an integer comparing two byte slices lexicographically.
    24  // The result will be 0 if a==b, -1 if a < b, and +1 if a > b.
    25  // A nil argument is equivalent to an empty slice.
    26  func Compare(a, b []byte) int {
    27  	return bytealg.Compare(a, b)
    28  }
    29  
    30  // explode splits s into a slice of UTF-8 sequences, one per Unicode code point (still slices of bytes),
    31  // up to a maximum of n byte slices. Invalid UTF-8 sequences are chopped into individual bytes.
    32  func explode(s []byte, n int) [][]byte {
    33  	if n <= 0 {
    34  		n = len(s)
    35  	}
    36  	a := make([][]byte, n)
    37  	var size int
    38  	na := 0
    39  	for len(s) > 0 {
    40  		if na+1 >= n {
    41  			a[na] = s
    42  			na++
    43  			break
    44  		}
    45  		_, size = utf8.DecodeRune(s)
    46  		a[na] = s[0:size:size]
    47  		s = s[size:]
    48  		na++
    49  	}
    50  	return a[0:na]
    51  }
    52  
    53  // Count counts the number of non-overlapping instances of sep in s.
    54  // If sep is an empty slice, Count returns 1 + the number of UTF-8-encoded code points in s.
    55  func Count(s, sep []byte) int {
    56  	// special case
    57  	if len(sep) == 0 {
    58  		return utf8.RuneCount(s) + 1
    59  	}
    60  	if len(sep) == 1 {
    61  		return bytealg.Count(s, sep[0])
    62  	}
    63  	n := 0
    64  	for {
    65  		i := Index(s, sep)
    66  		if i == -1 {
    67  			return n
    68  		}
    69  		n++
    70  		s = s[i+len(sep):]
    71  	}
    72  }
    73  
    74  // Contains reports whether subslice is within b.
    75  func Contains(b, subslice []byte) bool {
    76  	return Index(b, subslice) != -1
    77  }
    78  
    79  // ContainsAny reports whether any of the UTF-8-encoded code points in chars are within b.
    80  func ContainsAny(b []byte, chars string) bool {
    81  	return IndexAny(b, chars) >= 0
    82  }
    83  
    84  // ContainsRune reports whether the rune is contained in the UTF-8-encoded byte slice b.
    85  func ContainsRune(b []byte, r rune) bool {
    86  	return IndexRune(b, r) >= 0
    87  }
    88  
    89  // IndexByte returns the index of the first instance of c in b, or -1 if c is not present in b.
    90  func IndexByte(b []byte, c byte) int {
    91  	return bytealg.IndexByte(b, c)
    92  }
    93  
    94  func indexBytePortable(s []byte, c byte) int {
    95  	for i, b := range s {
    96  		if b == c {
    97  			return i
    98  		}
    99  	}
   100  	return -1
   101  }
   102  
   103  // LastIndex returns the index of the last instance of sep in s, or -1 if sep is not present in s.
   104  func LastIndex(s, sep []byte) int {
   105  	n := len(sep)
   106  	switch {
   107  	case n == 0:
   108  		return len(s)
   109  	case n == 1:
   110  		return LastIndexByte(s, sep[0])
   111  	case n == len(s):
   112  		if Equal(s, sep) {
   113  			return 0
   114  		}
   115  		return -1
   116  	case n > len(s):
   117  		return -1
   118  	}
   119  	// Rabin-Karp search from the end of the string
   120  	hashss, pow := bytealg.HashStrRevBytes(sep)
   121  	last := len(s) - n
   122  	var h uint32
   123  	for i := len(s) - 1; i >= last; i-- {
   124  		h = h*bytealg.PrimeRK + uint32(s[i])
   125  	}
   126  	if h == hashss && Equal(s[last:], sep) {
   127  		return last
   128  	}
   129  	for i := last - 1; i >= 0; i-- {
   130  		h *= bytealg.PrimeRK
   131  		h += uint32(s[i])
   132  		h -= pow * uint32(s[i+n])
   133  		if h == hashss && Equal(s[i:i+n], sep) {
   134  			return i
   135  		}
   136  	}
   137  	return -1
   138  }
   139  
   140  // LastIndexByte returns the index of the last instance of c in s, or -1 if c is not present in s.
   141  func LastIndexByte(s []byte, c byte) int {
   142  	for i := len(s) - 1; i >= 0; i-- {
   143  		if s[i] == c {
   144  			return i
   145  		}
   146  	}
   147  	return -1
   148  }
   149  
   150  // IndexRune interprets s as a sequence of UTF-8-encoded code points.
   151  // It returns the byte index of the first occurrence in s of the given rune.
   152  // It returns -1 if rune is not present in s.
   153  // If r is utf8.RuneError, it returns the first instance of any
   154  // invalid UTF-8 byte sequence.
   155  func IndexRune(s []byte, r rune) int {
   156  	switch {
   157  	case 0 <= r && r < utf8.RuneSelf:
   158  		return IndexByte(s, byte(r))
   159  	case r == utf8.RuneError:
   160  		for i := 0; i < len(s); {
   161  			r1, n := utf8.DecodeRune(s[i:])
   162  			if r1 == utf8.RuneError {
   163  				return i
   164  			}
   165  			i += n
   166  		}
   167  		return -1
   168  	case !utf8.ValidRune(r):
   169  		return -1
   170  	default:
   171  		var b [utf8.UTFMax]byte
   172  		n := utf8.EncodeRune(b[:], r)
   173  		return Index(s, b[:n])
   174  	}
   175  }
   176  
   177  // IndexAny interprets s as a sequence of UTF-8-encoded Unicode code points.
   178  // It returns the byte index of the first occurrence in s of any of the Unicode
   179  // code points in chars. It returns -1 if chars is empty or if there is no code
   180  // point in common.
   181  func IndexAny(s []byte, chars string) int {
   182  	if chars == "" {
   183  		// Avoid scanning all of s.
   184  		return -1
   185  	}
   186  	if len(s) == 1 {
   187  		r := rune(s[0])
   188  		if r >= utf8.RuneSelf {
   189  			// search utf8.RuneError.
   190  			for _, r = range chars {
   191  				if r == utf8.RuneError {
   192  					return 0
   193  				}
   194  			}
   195  			return -1
   196  		}
   197  		if bytealg.IndexByteString(chars, s[0]) >= 0 {
   198  			return 0
   199  		}
   200  		return -1
   201  	}
   202  	if len(chars) == 1 {
   203  		r := rune(chars[0])
   204  		if r >= utf8.RuneSelf {
   205  			r = utf8.RuneError
   206  		}
   207  		return IndexRune(s, r)
   208  	}
   209  	if len(s) > 8 {
   210  		if as, isASCII := makeASCIISet(chars); isASCII {
   211  			for i, c := range s {
   212  				if as.contains(c) {
   213  					return i
   214  				}
   215  			}
   216  			return -1
   217  		}
   218  	}
   219  	var width int
   220  	for i := 0; i < len(s); i += width {
   221  		r := rune(s[i])
   222  		if r < utf8.RuneSelf {
   223  			if bytealg.IndexByteString(chars, s[i]) >= 0 {
   224  				return i
   225  			}
   226  			width = 1
   227  			continue
   228  		}
   229  		r, width = utf8.DecodeRune(s[i:])
   230  		if r != utf8.RuneError {
   231  			// r is 2 to 4 bytes
   232  			if len(chars) == width {
   233  				if chars == string(r) {
   234  					return i
   235  				}
   236  				continue
   237  			}
   238  			// Use bytealg.IndexString for performance if available.
   239  			if bytealg.MaxLen >= width {
   240  				if bytealg.IndexString(chars, string(r)) >= 0 {
   241  					return i
   242  				}
   243  				continue
   244  			}
   245  		}
   246  		for _, ch := range chars {
   247  			if r == ch {
   248  				return i
   249  			}
   250  		}
   251  	}
   252  	return -1
   253  }
   254  
   255  // LastIndexAny interprets s as a sequence of UTF-8-encoded Unicode code
   256  // points. It returns the byte index of the last occurrence in s of any of
   257  // the Unicode code points in chars. It returns -1 if chars is empty or if
   258  // there is no code point in common.
   259  func LastIndexAny(s []byte, chars string) int {
   260  	if chars == "" {
   261  		// Avoid scanning all of s.
   262  		return -1
   263  	}
   264  	if len(s) > 8 {
   265  		if as, isASCII := makeASCIISet(chars); isASCII {
   266  			for i := len(s) - 1; i >= 0; i-- {
   267  				if as.contains(s[i]) {
   268  					return i
   269  				}
   270  			}
   271  			return -1
   272  		}
   273  	}
   274  	if len(s) == 1 {
   275  		r := rune(s[0])
   276  		if r >= utf8.RuneSelf {
   277  			for _, r = range chars {
   278  				if r == utf8.RuneError {
   279  					return 0
   280  				}
   281  			}
   282  			return -1
   283  		}
   284  		if bytealg.IndexByteString(chars, s[0]) >= 0 {
   285  			return 0
   286  		}
   287  		return -1
   288  	}
   289  	if len(chars) == 1 {
   290  		cr := rune(chars[0])
   291  		if cr >= utf8.RuneSelf {
   292  			cr = utf8.RuneError
   293  		}
   294  		for i := len(s); i > 0; {
   295  			r, size := utf8.DecodeLastRune(s[:i])
   296  			i -= size
   297  			if r == cr {
   298  				return i
   299  			}
   300  		}
   301  		return -1
   302  	}
   303  	for i := len(s); i > 0; {
   304  		r := rune(s[i-1])
   305  		if r < utf8.RuneSelf {
   306  			if bytealg.IndexByteString(chars, s[i-1]) >= 0 {
   307  				return i - 1
   308  			}
   309  			i--
   310  			continue
   311  		}
   312  		r, size := utf8.DecodeLastRune(s[:i])
   313  		i -= size
   314  		if r != utf8.RuneError {
   315  			// r is 2 to 4 bytes
   316  			if len(chars) == size {
   317  				if chars == string(r) {
   318  					return i
   319  				}
   320  				continue
   321  			}
   322  			// Use bytealg.IndexString for performance if available.
   323  			if bytealg.MaxLen >= size {
   324  				if bytealg.IndexString(chars, string(r)) >= 0 {
   325  					return i
   326  				}
   327  				continue
   328  			}
   329  		}
   330  		for _, ch := range chars {
   331  			if r == ch {
   332  				return i
   333  			}
   334  		}
   335  	}
   336  	return -1
   337  }
   338  
   339  // Generic split: splits after each instance of sep,
   340  // including sepSave bytes of sep in the subslices.
   341  func genSplit(s, sep []byte, sepSave, n int) [][]byte {
   342  	if n == 0 {
   343  		return nil
   344  	}
   345  	if len(sep) == 0 {
   346  		return explode(s, n)
   347  	}
   348  	if n < 0 {
   349  		n = Count(s, sep) + 1
   350  	}
   351  
   352  	a := make([][]byte, n)
   353  	n--
   354  	i := 0
   355  	for i < n {
   356  		m := Index(s, sep)
   357  		if m < 0 {
   358  			break
   359  		}
   360  		a[i] = s[: m+sepSave : m+sepSave]
   361  		s = s[m+len(sep):]
   362  		i++
   363  	}
   364  	a[i] = s
   365  	return a[:i+1]
   366  }
   367  
   368  // SplitN slices s into subslices separated by sep and returns a slice of
   369  // the subslices between those separators.
   370  // If sep is empty, SplitN splits after each UTF-8 sequence.
   371  // The count determines the number of subslices to return:
   372  //   n > 0: at most n subslices; the last subslice will be the unsplit remainder.
   373  //   n == 0: the result is nil (zero subslices)
   374  //   n < 0: all subslices
   375  func SplitN(s, sep []byte, n int) [][]byte { return genSplit(s, sep, 0, n) }
   376  
   377  // SplitAfterN slices s into subslices after each instance of sep and
   378  // returns a slice of those subslices.
   379  // If sep is empty, SplitAfterN splits after each UTF-8 sequence.
   380  // The count determines the number of subslices to return:
   381  //   n > 0: at most n subslices; the last subslice will be the unsplit remainder.
   382  //   n == 0: the result is nil (zero subslices)
   383  //   n < 0: all subslices
   384  func SplitAfterN(s, sep []byte, n int) [][]byte {
   385  	return genSplit(s, sep, len(sep), n)
   386  }
   387  
   388  // Split slices s into all subslices separated by sep and returns a slice of
   389  // the subslices between those separators.
   390  // If sep is empty, Split splits after each UTF-8 sequence.
   391  // It is equivalent to SplitN with a count of -1.
   392  func Split(s, sep []byte) [][]byte { return genSplit(s, sep, 0, -1) }
   393  
   394  // SplitAfter slices s into all subslices after each instance of sep and
   395  // returns a slice of those subslices.
   396  // If sep is empty, SplitAfter splits after each UTF-8 sequence.
   397  // It is equivalent to SplitAfterN with a count of -1.
   398  func SplitAfter(s, sep []byte) [][]byte {
   399  	return genSplit(s, sep, len(sep), -1)
   400  }
   401  
   402  var asciiSpace = [256]uint8{'\t': 1, '\n': 1, '\v': 1, '\f': 1, '\r': 1, ' ': 1}
   403  
   404  // Fields interprets s as a sequence of UTF-8-encoded code points.
   405  // It splits the slice s around each instance of one or more consecutive white space
   406  // characters, as defined by unicode.IsSpace, returning a slice of subslices of s or an
   407  // empty slice if s contains only white space.
   408  func Fields(s []byte) [][]byte {
   409  	// First count the fields.
   410  	// This is an exact count if s is ASCII, otherwise it is an approximation.
   411  	n := 0
   412  	wasSpace := 1
   413  	// setBits is used to track which bits are set in the bytes of s.
   414  	setBits := uint8(0)
   415  	for i := 0; i < len(s); i++ {
   416  		r := s[i]
   417  		setBits |= r
   418  		isSpace := int(asciiSpace[r])
   419  		n += wasSpace & ^isSpace
   420  		wasSpace = isSpace
   421  	}
   422  
   423  	if setBits >= utf8.RuneSelf {
   424  		// Some runes in the input slice are not ASCII.
   425  		return FieldsFunc(s, unicode.IsSpace)
   426  	}
   427  
   428  	// ASCII fast path
   429  	a := make([][]byte, n)
   430  	na := 0
   431  	fieldStart := 0
   432  	i := 0
   433  	// Skip spaces in the front of the input.
   434  	for i < len(s) && asciiSpace[s[i]] != 0 {
   435  		i++
   436  	}
   437  	fieldStart = i
   438  	for i < len(s) {
   439  		if asciiSpace[s[i]] == 0 {
   440  			i++
   441  			continue
   442  		}
   443  		a[na] = s[fieldStart:i:i]
   444  		na++
   445  		i++
   446  		// Skip spaces in between fields.
   447  		for i < len(s) && asciiSpace[s[i]] != 0 {
   448  			i++
   449  		}
   450  		fieldStart = i
   451  	}
   452  	if fieldStart < len(s) { // Last field might end at EOF.
   453  		a[na] = s[fieldStart:len(s):len(s)]
   454  	}
   455  	return a
   456  }
   457  
   458  // FieldsFunc interprets s as a sequence of UTF-8-encoded code points.
   459  // It splits the slice s at each run of code points c satisfying f(c) and
   460  // returns a slice of subslices of s. If all code points in s satisfy f(c), or
   461  // len(s) == 0, an empty slice is returned.
   462  //
   463  // FieldsFunc makes no guarantees about the order in which it calls f(c)
   464  // and assumes that f always returns the same value for a given c.
   465  func FieldsFunc(s []byte, f func(rune) bool) [][]byte {
   466  	// A span is used to record a slice of s of the form s[start:end].
   467  	// The start index is inclusive and the end index is exclusive.
   468  	type span struct {
   469  		start int
   470  		end   int
   471  	}
   472  	spans := make([]span, 0, 32)
   473  
   474  	// Find the field start and end indices.
   475  	// Doing this in a separate pass (rather than slicing the string s
   476  	// and collecting the result substrings right away) is significantly
   477  	// more efficient, possibly due to cache effects.
   478  	start := -1 // valid span start if >= 0
   479  	for i := 0; i < len(s); {
   480  		size := 1
   481  		r := rune(s[i])
   482  		if r >= utf8.RuneSelf {
   483  			r, size = utf8.DecodeRune(s[i:])
   484  		}
   485  		if f(r) {
   486  			if start >= 0 {
   487  				spans = append(spans, span{start, i})
   488  				start = -1
   489  			}
   490  		} else {
   491  			if start < 0 {
   492  				start = i
   493  			}
   494  		}
   495  		i += size
   496  	}
   497  
   498  	// Last field might end at EOF.
   499  	if start >= 0 {
   500  		spans = append(spans, span{start, len(s)})
   501  	}
   502  
   503  	// Create subslices from recorded field indices.
   504  	a := make([][]byte, len(spans))
   505  	for i, span := range spans {
   506  		a[i] = s[span.start:span.end:span.end]
   507  	}
   508  
   509  	return a
   510  }
   511  
   512  // Join concatenates the elements of s to create a new byte slice. The separator
   513  // sep is placed between elements in the resulting slice.
   514  func Join(s [][]byte, sep []byte) []byte {
   515  	if len(s) == 0 {
   516  		return []byte{}
   517  	}
   518  	if len(s) == 1 {
   519  		// Just return a copy.
   520  		return append([]byte(nil), s[0]...)
   521  	}
   522  	n := len(sep) * (len(s) - 1)
   523  	for _, v := range s {
   524  		n += len(v)
   525  	}
   526  
   527  	b := make([]byte, n)
   528  	bp := copy(b, s[0])
   529  	for _, v := range s[1:] {
   530  		bp += copy(b[bp:], sep)
   531  		bp += copy(b[bp:], v)
   532  	}
   533  	return b
   534  }
   535  
   536  // HasPrefix tests whether the byte slice s begins with prefix.
   537  func HasPrefix(s, prefix []byte) bool {
   538  	return len(s) >= len(prefix) && Equal(s[0:len(prefix)], prefix)
   539  }
   540  
   541  // HasSuffix tests whether the byte slice s ends with suffix.
   542  func HasSuffix(s, suffix []byte) bool {
   543  	return len(s) >= len(suffix) && Equal(s[len(s)-len(suffix):], suffix)
   544  }
   545  
   546  // Map returns a copy of the byte slice s with all its characters modified
   547  // according to the mapping function. If mapping returns a negative value, the character is
   548  // dropped from the byte slice with no replacement. The characters in s and the
   549  // output are interpreted as UTF-8-encoded code points.
   550  func Map(mapping func(r rune) rune, s []byte) []byte {
   551  	// In the worst case, the slice can grow when mapped, making
   552  	// things unpleasant. But it's so rare we barge in assuming it's
   553  	// fine. It could also shrink but that falls out naturally.
   554  	maxbytes := len(s) // length of b
   555  	nbytes := 0        // number of bytes encoded in b
   556  	b := make([]byte, maxbytes)
   557  	for i := 0; i < len(s); {
   558  		wid := 1
   559  		r := rune(s[i])
   560  		if r >= utf8.RuneSelf {
   561  			r, wid = utf8.DecodeRune(s[i:])
   562  		}
   563  		r = mapping(r)
   564  		if r >= 0 {
   565  			rl := utf8.RuneLen(r)
   566  			if rl < 0 {
   567  				rl = len(string(utf8.RuneError))
   568  			}
   569  			if nbytes+rl > maxbytes {
   570  				// Grow the buffer.
   571  				maxbytes = maxbytes*2 + utf8.UTFMax
   572  				nb := make([]byte, maxbytes)
   573  				copy(nb, b[0:nbytes])
   574  				b = nb
   575  			}
   576  			nbytes += utf8.EncodeRune(b[nbytes:maxbytes], r)
   577  		}
   578  		i += wid
   579  	}
   580  	return b[0:nbytes]
   581  }
   582  
   583  // Repeat returns a new byte slice consisting of count copies of b.
   584  //
   585  // It panics if count is negative or if
   586  // the result of (len(b) * count) overflows.
   587  func Repeat(b []byte, count int) []byte {
   588  	if count == 0 {
   589  		return []byte{}
   590  	}
   591  	// Since we cannot return an error on overflow,
   592  	// we should panic if the repeat will generate
   593  	// an overflow.
   594  	// See Issue golang.org/issue/16237.
   595  	if count < 0 {
   596  		panic("bytes: negative Repeat count")
   597  	} else if len(b)*count/count != len(b) {
   598  		panic("bytes: Repeat count causes overflow")
   599  	}
   600  
   601  	nb := make([]byte, len(b)*count)
   602  	bp := copy(nb, b)
   603  	for bp < len(nb) {
   604  		copy(nb[bp:], nb[:bp])
   605  		bp *= 2
   606  	}
   607  	return nb
   608  }
   609  
   610  // ToUpper returns a copy of the byte slice s with all Unicode letters mapped to
   611  // their upper case.
   612  func ToUpper(s []byte) []byte {
   613  	isASCII, hasLower := true, false
   614  	for i := 0; i < len(s); i++ {
   615  		c := s[i]
   616  		if c >= utf8.RuneSelf {
   617  			isASCII = false
   618  			break
   619  		}
   620  		hasLower = hasLower || ('a' <= c && c <= 'z')
   621  	}
   622  
   623  	if isASCII { // optimize for ASCII-only byte slices.
   624  		if !hasLower {
   625  			// Just return a copy.
   626  			return append([]byte(""), s...)
   627  		}
   628  		b := make([]byte, len(s))
   629  		for i := 0; i < len(s); i++ {
   630  			c := s[i]
   631  			if 'a' <= c && c <= 'z' {
   632  				c -= 'a' - 'A'
   633  			}
   634  			b[i] = c
   635  		}
   636  		return b
   637  	}
   638  	return Map(unicode.ToUpper, s)
   639  }
   640  
   641  // ToLower returns a copy of the byte slice s with all Unicode letters mapped to
   642  // their lower case.
   643  func ToLower(s []byte) []byte {
   644  	isASCII, hasUpper := true, false
   645  	for i := 0; i < len(s); i++ {
   646  		c := s[i]
   647  		if c >= utf8.RuneSelf {
   648  			isASCII = false
   649  			break
   650  		}
   651  		hasUpper = hasUpper || ('A' <= c && c <= 'Z')
   652  	}
   653  
   654  	if isASCII { // optimize for ASCII-only byte slices.
   655  		if !hasUpper {
   656  			return append([]byte(""), s...)
   657  		}
   658  		b := make([]byte, len(s))
   659  		for i := 0; i < len(s); i++ {
   660  			c := s[i]
   661  			if 'A' <= c && c <= 'Z' {
   662  				c += 'a' - 'A'
   663  			}
   664  			b[i] = c
   665  		}
   666  		return b
   667  	}
   668  	return Map(unicode.ToLower, s)
   669  }
   670  
   671  // ToTitle treats s as UTF-8-encoded bytes and returns a copy with all the Unicode letters mapped to their title case.
   672  func ToTitle(s []byte) []byte { return Map(unicode.ToTitle, s) }
   673  
   674  // ToUpperSpecial treats s as UTF-8-encoded bytes and returns a copy with all the Unicode letters mapped to their
   675  // upper case, giving priority to the special casing rules.
   676  func ToUpperSpecial(c unicode.SpecialCase, s []byte) []byte {
   677  	return Map(c.ToUpper, s)
   678  }
   679  
   680  // ToLowerSpecial treats s as UTF-8-encoded bytes and returns a copy with all the Unicode letters mapped to their
   681  // lower case, giving priority to the special casing rules.
   682  func ToLowerSpecial(c unicode.SpecialCase, s []byte) []byte {
   683  	return Map(c.ToLower, s)
   684  }
   685  
   686  // ToTitleSpecial treats s as UTF-8-encoded bytes and returns a copy with all the Unicode letters mapped to their
   687  // title case, giving priority to the special casing rules.
   688  func ToTitleSpecial(c unicode.SpecialCase, s []byte) []byte {
   689  	return Map(c.ToTitle, s)
   690  }
   691  
   692  // ToValidUTF8 treats s as UTF-8-encoded bytes and returns a copy with each run of bytes
   693  // representing invalid UTF-8 replaced with the bytes in replacement, which may be empty.
   694  func ToValidUTF8(s, replacement []byte) []byte {
   695  	b := make([]byte, 0, len(s)+len(replacement))
   696  	invalid := false // previous byte was from an invalid UTF-8 sequence
   697  	for i := 0; i < len(s); {
   698  		c := s[i]
   699  		if c < utf8.RuneSelf {
   700  			i++
   701  			invalid = false
   702  			b = append(b, byte(c))
   703  			continue
   704  		}
   705  		_, wid := utf8.DecodeRune(s[i:])
   706  		if wid == 1 {
   707  			i++
   708  			if !invalid {
   709  				invalid = true
   710  				b = append(b, replacement...)
   711  			}
   712  			continue
   713  		}
   714  		invalid = false
   715  		b = append(b, s[i:i+wid]...)
   716  		i += wid
   717  	}
   718  	return b
   719  }
   720  
   721  // isSeparator reports whether the rune could mark a word boundary.
   722  // TODO: update when package unicode captures more of the properties.
   723  func isSeparator(r rune) bool {
   724  	// ASCII alphanumerics and underscore are not separators
   725  	if r <= 0x7F {
   726  		switch {
   727  		case '0' <= r && r <= '9':
   728  			return false
   729  		case 'a' <= r && r <= 'z':
   730  			return false
   731  		case 'A' <= r && r <= 'Z':
   732  			return false
   733  		case r == '_':
   734  			return false
   735  		}
   736  		return true
   737  	}
   738  	// Letters and digits are not separators
   739  	if unicode.IsLetter(r) || unicode.IsDigit(r) {
   740  		return false
   741  	}
   742  	// Otherwise, all we can do for now is treat spaces as separators.
   743  	return unicode.IsSpace(r)
   744  }
   745  
   746  // Title treats s as UTF-8-encoded bytes and returns a copy with all Unicode letters that begin
   747  // words mapped to their title case.
   748  //
   749  // BUG(rsc): The rule Title uses for word boundaries does not handle Unicode punctuation properly.
   750  func Title(s []byte) []byte {
   751  	// Use a closure here to remember state.
   752  	// Hackish but effective. Depends on Map scanning in order and calling
   753  	// the closure once per rune.
   754  	prev := ' '
   755  	return Map(
   756  		func(r rune) rune {
   757  			if isSeparator(prev) {
   758  				prev = r
   759  				return unicode.ToTitle(r)
   760  			}
   761  			prev = r
   762  			return r
   763  		},
   764  		s)
   765  }
   766  
   767  // TrimLeftFunc treats s as UTF-8-encoded bytes and returns a subslice of s by slicing off
   768  // all leading UTF-8-encoded code points c that satisfy f(c).
   769  func TrimLeftFunc(s []byte, f func(r rune) bool) []byte {
   770  	i := indexFunc(s, f, false)
   771  	if i == -1 {
   772  		return nil
   773  	}
   774  	return s[i:]
   775  }
   776  
   777  // TrimRightFunc returns a subslice of s by slicing off all trailing
   778  // UTF-8-encoded code points c that satisfy f(c).
   779  func TrimRightFunc(s []byte, f func(r rune) bool) []byte {
   780  	i := lastIndexFunc(s, f, false)
   781  	if i >= 0 && s[i] >= utf8.RuneSelf {
   782  		_, wid := utf8.DecodeRune(s[i:])
   783  		i += wid
   784  	} else {
   785  		i++
   786  	}
   787  	return s[0:i]
   788  }
   789  
   790  // TrimFunc returns a subslice of s by slicing off all leading and trailing
   791  // UTF-8-encoded code points c that satisfy f(c).
   792  func TrimFunc(s []byte, f func(r rune) bool) []byte {
   793  	return TrimRightFunc(TrimLeftFunc(s, f), f)
   794  }
   795  
   796  // TrimPrefix returns s without the provided leading prefix string.
   797  // If s doesn't start with prefix, s is returned unchanged.
   798  func TrimPrefix(s, prefix []byte) []byte {
   799  	if HasPrefix(s, prefix) {
   800  		return s[len(prefix):]
   801  	}
   802  	return s
   803  }
   804  
   805  // TrimSuffix returns s without the provided trailing suffix string.
   806  // If s doesn't end with suffix, s is returned unchanged.
   807  func TrimSuffix(s, suffix []byte) []byte {
   808  	if HasSuffix(s, suffix) {
   809  		return s[:len(s)-len(suffix)]
   810  	}
   811  	return s
   812  }
   813  
   814  // IndexFunc interprets s as a sequence of UTF-8-encoded code points.
   815  // It returns the byte index in s of the first Unicode
   816  // code point satisfying f(c), or -1 if none do.
   817  func IndexFunc(s []byte, f func(r rune) bool) int {
   818  	return indexFunc(s, f, true)
   819  }
   820  
   821  // LastIndexFunc interprets s as a sequence of UTF-8-encoded code points.
   822  // It returns the byte index in s of the last Unicode
   823  // code point satisfying f(c), or -1 if none do.
   824  func LastIndexFunc(s []byte, f func(r rune) bool) int {
   825  	return lastIndexFunc(s, f, true)
   826  }
   827  
   828  // indexFunc is the same as IndexFunc except that if
   829  // truth==false, the sense of the predicate function is
   830  // inverted.
   831  func indexFunc(s []byte, f func(r rune) bool, truth bool) int {
   832  	start := 0
   833  	for start < len(s) {
   834  		wid := 1
   835  		r := rune(s[start])
   836  		if r >= utf8.RuneSelf {
   837  			r, wid = utf8.DecodeRune(s[start:])
   838  		}
   839  		if f(r) == truth {
   840  			return start
   841  		}
   842  		start += wid
   843  	}
   844  	return -1
   845  }
   846  
   847  // lastIndexFunc is the same as LastIndexFunc except that if
   848  // truth==false, the sense of the predicate function is
   849  // inverted.
   850  func lastIndexFunc(s []byte, f func(r rune) bool, truth bool) int {
   851  	for i := len(s); i > 0; {
   852  		r, size := rune(s[i-1]), 1
   853  		if r >= utf8.RuneSelf {
   854  			r, size = utf8.DecodeLastRune(s[0:i])
   855  		}
   856  		i -= size
   857  		if f(r) == truth {
   858  			return i
   859  		}
   860  	}
   861  	return -1
   862  }
   863  
   864  // asciiSet is a 32-byte value, where each bit represents the presence of a
   865  // given ASCII character in the set. The 128-bits of the lower 16 bytes,
   866  // starting with the least-significant bit of the lowest word to the
   867  // most-significant bit of the highest word, map to the full range of all
   868  // 128 ASCII characters. The 128-bits of the upper 16 bytes will be zeroed,
   869  // ensuring that any non-ASCII character will be reported as not in the set.
   870  type asciiSet [8]uint32
   871  
   872  // makeASCIISet creates a set of ASCII characters and reports whether all
   873  // characters in chars are ASCII.
   874  func makeASCIISet(chars string) (as asciiSet, ok bool) {
   875  	for i := 0; i < len(chars); i++ {
   876  		c := chars[i]
   877  		if c >= utf8.RuneSelf {
   878  			return as, false
   879  		}
   880  		as[c>>5] |= 1 << uint(c&31)
   881  	}
   882  	return as, true
   883  }
   884  
   885  // contains reports whether c is inside the set.
   886  func (as *asciiSet) contains(c byte) bool {
   887  	return (as[c>>5] & (1 << uint(c&31))) != 0
   888  }
   889  
   890  func makeCutsetFunc(cutset string) func(r rune) bool {
   891  	if len(cutset) == 1 && cutset[0] < utf8.RuneSelf {
   892  		return func(r rune) bool {
   893  			return r == rune(cutset[0])
   894  		}
   895  	}
   896  	if as, isASCII := makeASCIISet(cutset); isASCII {
   897  		return func(r rune) bool {
   898  			return r < utf8.RuneSelf && as.contains(byte(r))
   899  		}
   900  	}
   901  	return func(r rune) bool {
   902  		for _, c := range cutset {
   903  			if c == r {
   904  				return true
   905  			}
   906  		}
   907  		return false
   908  	}
   909  }
   910  
   911  // Trim returns a subslice of s by slicing off all leading and
   912  // trailing UTF-8-encoded code points contained in cutset.
   913  func Trim(s []byte, cutset string) []byte {
   914  	return TrimFunc(s, makeCutsetFunc(cutset))
   915  }
   916  
   917  // TrimLeft returns a subslice of s by slicing off all leading
   918  // UTF-8-encoded code points contained in cutset.
   919  func TrimLeft(s []byte, cutset string) []byte {
   920  	return TrimLeftFunc(s, makeCutsetFunc(cutset))
   921  }
   922  
   923  // TrimRight returns a subslice of s by slicing off all trailing
   924  // UTF-8-encoded code points that are contained in cutset.
   925  func TrimRight(s []byte, cutset string) []byte {
   926  	return TrimRightFunc(s, makeCutsetFunc(cutset))
   927  }
   928  
   929  // TrimSpace returns a subslice of s by slicing off all leading and
   930  // trailing white space, as defined by Unicode.
   931  func TrimSpace(s []byte) []byte {
   932  	// Fast path for ASCII: look for the first ASCII non-space byte
   933  	start := 0
   934  	for ; start < len(s); start++ {
   935  		c := s[start]
   936  		if c >= utf8.RuneSelf {
   937  			// If we run into a non-ASCII byte, fall back to the
   938  			// slower unicode-aware method on the remaining bytes
   939  			return TrimFunc(s[start:], unicode.IsSpace)
   940  		}
   941  		if asciiSpace[c] == 0 {
   942  			break
   943  		}
   944  	}
   945  
   946  	// Now look for the first ASCII non-space byte from the end
   947  	stop := len(s)
   948  	for ; stop > start; stop-- {
   949  		c := s[stop-1]
   950  		if c >= utf8.RuneSelf {
   951  			return TrimFunc(s[start:stop], unicode.IsSpace)
   952  		}
   953  		if asciiSpace[c] == 0 {
   954  			break
   955  		}
   956  	}
   957  
   958  	// At this point s[start:stop] starts and ends with an ASCII
   959  	// non-space bytes, so we're done. Non-ASCII cases have already
   960  	// been handled above.
   961  	if start == stop {
   962  		// Special case to preserve previous TrimLeftFunc behavior,
   963  		// returning nil instead of empty slice if all spaces.
   964  		return nil
   965  	}
   966  	return s[start:stop]
   967  }
   968  
   969  // Runes interprets s as a sequence of UTF-8-encoded code points.
   970  // It returns a slice of runes (Unicode code points) equivalent to s.
   971  func Runes(s []byte) []rune {
   972  	t := make([]rune, utf8.RuneCount(s))
   973  	i := 0
   974  	for len(s) > 0 {
   975  		r, l := utf8.DecodeRune(s)
   976  		t[i] = r
   977  		i++
   978  		s = s[l:]
   979  	}
   980  	return t
   981  }
   982  
   983  // Replace returns a copy of the slice s with the first n
   984  // non-overlapping instances of old replaced by new.
   985  // If old is empty, it matches at the beginning of the slice
   986  // and after each UTF-8 sequence, yielding up to k+1 replacements
   987  // for a k-rune slice.
   988  // If n < 0, there is no limit on the number of replacements.
   989  func Replace(s, old, new []byte, n int) []byte {
   990  	m := 0
   991  	if n != 0 {
   992  		// Compute number of replacements.
   993  		m = Count(s, old)
   994  	}
   995  	if m == 0 {
   996  		// Just return a copy.
   997  		return append([]byte(nil), s...)
   998  	}
   999  	if n < 0 || m < n {
  1000  		n = m
  1001  	}
  1002  
  1003  	// Apply replacements to buffer.
  1004  	t := make([]byte, len(s)+n*(len(new)-len(old)))
  1005  	w := 0
  1006  	start := 0
  1007  	for i := 0; i < n; i++ {
  1008  		j := start
  1009  		if len(old) == 0 {
  1010  			if i > 0 {
  1011  				_, wid := utf8.DecodeRune(s[start:])
  1012  				j += wid
  1013  			}
  1014  		} else {
  1015  			j += Index(s[start:], old)
  1016  		}
  1017  		w += copy(t[w:], s[start:j])
  1018  		w += copy(t[w:], new)
  1019  		start = j + len(old)
  1020  	}
  1021  	w += copy(t[w:], s[start:])
  1022  	return t[0:w]
  1023  }
  1024  
  1025  // ReplaceAll returns a copy of the slice s with all
  1026  // non-overlapping instances of old replaced by new.
  1027  // If old is empty, it matches at the beginning of the slice
  1028  // and after each UTF-8 sequence, yielding up to k+1 replacements
  1029  // for a k-rune slice.
  1030  func ReplaceAll(s, old, new []byte) []byte {
  1031  	return Replace(s, old, new, -1)
  1032  }
  1033  
  1034  // EqualFold reports whether s and t, interpreted as UTF-8 strings,
  1035  // are equal under Unicode case-folding, which is a more general
  1036  // form of case-insensitivity.
  1037  func EqualFold(s, t []byte) bool {
  1038  	for len(s) != 0 && len(t) != 0 {
  1039  		// Extract first rune from each.
  1040  		var sr, tr rune
  1041  		if s[0] < utf8.RuneSelf {
  1042  			sr, s = rune(s[0]), s[1:]
  1043  		} else {
  1044  			r, size := utf8.DecodeRune(s)
  1045  			sr, s = r, s[size:]
  1046  		}
  1047  		if t[0] < utf8.RuneSelf {
  1048  			tr, t = rune(t[0]), t[1:]
  1049  		} else {
  1050  			r, size := utf8.DecodeRune(t)
  1051  			tr, t = r, t[size:]
  1052  		}
  1053  
  1054  		// If they match, keep going; if not, return false.
  1055  
  1056  		// Easy case.
  1057  		if tr == sr {
  1058  			continue
  1059  		}
  1060  
  1061  		// Make sr < tr to simplify what follows.
  1062  		if tr < sr {
  1063  			tr, sr = sr, tr
  1064  		}
  1065  		// Fast check for ASCII.
  1066  		if tr < utf8.RuneSelf {
  1067  			// ASCII only, sr/tr must be upper/lower case
  1068  			if 'A' <= sr && sr <= 'Z' && tr == sr+'a'-'A' {
  1069  				continue
  1070  			}
  1071  			return false
  1072  		}
  1073  
  1074  		// General case. SimpleFold(x) returns the next equivalent rune > x
  1075  		// or wraps around to smaller values.
  1076  		r := unicode.SimpleFold(sr)
  1077  		for r != sr && r < tr {
  1078  			r = unicode.SimpleFold(r)
  1079  		}
  1080  		if r == tr {
  1081  			continue
  1082  		}
  1083  		return false
  1084  	}
  1085  
  1086  	// One string is empty. Are both?
  1087  	return len(s) == len(t)
  1088  }
  1089  
  1090  // Index returns the index of the first instance of sep in s, or -1 if sep is not present in s.
  1091  func Index(s, sep []byte) int {
  1092  	n := len(sep)
  1093  	switch {
  1094  	case n == 0:
  1095  		return 0
  1096  	case n == 1:
  1097  		return IndexByte(s, sep[0])
  1098  	case n == len(s):
  1099  		if Equal(sep, s) {
  1100  			return 0
  1101  		}
  1102  		return -1
  1103  	case n > len(s):
  1104  		return -1
  1105  	case n <= bytealg.MaxLen:
  1106  		// Use brute force when s and sep both are small
  1107  		if len(s) <= bytealg.MaxBruteForce {
  1108  			return bytealg.Index(s, sep)
  1109  		}
  1110  		c0 := sep[0]
  1111  		c1 := sep[1]
  1112  		i := 0
  1113  		t := len(s) - n + 1
  1114  		fails := 0
  1115  		for i < t {
  1116  			if s[i] != c0 {
  1117  				// IndexByte is faster than bytealg.Index, so use it as long as
  1118  				// we're not getting lots of false positives.
  1119  				o := IndexByte(s[i+1:t], c0)
  1120  				if o < 0 {
  1121  					return -1
  1122  				}
  1123  				i += o + 1
  1124  			}
  1125  			if s[i+1] == c1 && Equal(s[i:i+n], sep) {
  1126  				return i
  1127  			}
  1128  			fails++
  1129  			i++
  1130  			// Switch to bytealg.Index when IndexByte produces too many false positives.
  1131  			if fails > bytealg.Cutover(i) {
  1132  				r := bytealg.Index(s[i:], sep)
  1133  				if r >= 0 {
  1134  					return r + i
  1135  				}
  1136  				return -1
  1137  			}
  1138  		}
  1139  		return -1
  1140  	}
  1141  	c0 := sep[0]
  1142  	c1 := sep[1]
  1143  	i := 0
  1144  	fails := 0
  1145  	t := len(s) - n + 1
  1146  	for i < t {
  1147  		if s[i] != c0 {
  1148  			o := IndexByte(s[i+1:t], c0)
  1149  			if o < 0 {
  1150  				break
  1151  			}
  1152  			i += o + 1
  1153  		}
  1154  		if s[i+1] == c1 && Equal(s[i:i+n], sep) {
  1155  			return i
  1156  		}
  1157  		i++
  1158  		fails++
  1159  		if fails >= 4+i>>4 && i < t {
  1160  			// Give up on IndexByte, it isn't skipping ahead
  1161  			// far enough to be better than Rabin-Karp.
  1162  			// Experiments (using IndexPeriodic) suggest
  1163  			// the cutover is about 16 byte skips.
  1164  			// TODO: if large prefixes of sep are matching
  1165  			// we should cutover at even larger average skips,
  1166  			// because Equal becomes that much more expensive.
  1167  			// This code does not take that effect into account.
  1168  			j := bytealg.IndexRabinKarpBytes(s[i:], sep)
  1169  			if j < 0 {
  1170  				return -1
  1171  			}
  1172  			return i + j
  1173  		}
  1174  	}
  1175  	return -1
  1176  }
  1177  

View as plain text