...

Source file src/strings/strings.go

Documentation: strings

		 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 strings implements simple functions to manipulate UTF-8 encoded strings.
		 6  //
		 7  // For information about UTF-8 strings in Go, see https://blog.golang.org/strings.
		 8  package strings
		 9  
		10  import (
		11  	"internal/bytealg"
		12  	"unicode"
		13  	"unicode/utf8"
		14  )
		15  
		16  // explode splits s into a slice of UTF-8 strings,
		17  // one string per Unicode character up to a maximum of n (n < 0 means no limit).
		18  // Invalid UTF-8 sequences become correct encodings of U+FFFD.
		19  func explode(s string, n int) []string {
		20  	l := utf8.RuneCountInString(s)
		21  	if n < 0 || n > l {
		22  		n = l
		23  	}
		24  	a := make([]string, n)
		25  	for i := 0; i < n-1; i++ {
		26  		ch, size := utf8.DecodeRuneInString(s)
		27  		a[i] = s[:size]
		28  		s = s[size:]
		29  		if ch == utf8.RuneError {
		30  			a[i] = string(utf8.RuneError)
		31  		}
		32  	}
		33  	if n > 0 {
		34  		a[n-1] = s
		35  	}
		36  	return a
		37  }
		38  
		39  // Count counts the number of non-overlapping instances of substr in s.
		40  // If substr is an empty string, Count returns 1 + the number of Unicode code points in s.
		41  func Count(s, substr string) int {
		42  	// special case
		43  	if len(substr) == 0 {
		44  		return utf8.RuneCountInString(s) + 1
		45  	}
		46  	if len(substr) == 1 {
		47  		return bytealg.CountString(s, substr[0])
		48  	}
		49  	n := 0
		50  	for {
		51  		i := Index(s, substr)
		52  		if i == -1 {
		53  			return n
		54  		}
		55  		n++
		56  		s = s[i+len(substr):]
		57  	}
		58  }
		59  
		60  // Contains reports whether substr is within s.
		61  func Contains(s, substr string) bool {
		62  	return Index(s, substr) >= 0
		63  }
		64  
		65  // ContainsAny reports whether any Unicode code points in chars are within s.
		66  func ContainsAny(s, chars string) bool {
		67  	return IndexAny(s, chars) >= 0
		68  }
		69  
		70  // ContainsRune reports whether the Unicode code point r is within s.
		71  func ContainsRune(s string, r rune) bool {
		72  	return IndexRune(s, r) >= 0
		73  }
		74  
		75  // LastIndex returns the index of the last instance of substr in s, or -1 if substr is not present in s.
		76  func LastIndex(s, substr string) int {
		77  	n := len(substr)
		78  	switch {
		79  	case n == 0:
		80  		return len(s)
		81  	case n == 1:
		82  		return LastIndexByte(s, substr[0])
		83  	case n == len(s):
		84  		if substr == s {
		85  			return 0
		86  		}
		87  		return -1
		88  	case n > len(s):
		89  		return -1
		90  	}
		91  	// Rabin-Karp search from the end of the string
		92  	hashss, pow := bytealg.HashStrRev(substr)
		93  	last := len(s) - n
		94  	var h uint32
		95  	for i := len(s) - 1; i >= last; i-- {
		96  		h = h*bytealg.PrimeRK + uint32(s[i])
		97  	}
		98  	if h == hashss && s[last:] == substr {
		99  		return last
	 100  	}
	 101  	for i := last - 1; i >= 0; i-- {
	 102  		h *= bytealg.PrimeRK
	 103  		h += uint32(s[i])
	 104  		h -= pow * uint32(s[i+n])
	 105  		if h == hashss && s[i:i+n] == substr {
	 106  			return i
	 107  		}
	 108  	}
	 109  	return -1
	 110  }
	 111  
	 112  // IndexByte returns the index of the first instance of c in s, or -1 if c is not present in s.
	 113  func IndexByte(s string, c byte) int {
	 114  	return bytealg.IndexByteString(s, c)
	 115  }
	 116  
	 117  // IndexRune returns the index of the first instance of the Unicode code point
	 118  // r, or -1 if rune is not present in s.
	 119  // If r is utf8.RuneError, it returns the first instance of any
	 120  // invalid UTF-8 byte sequence.
	 121  func IndexRune(s string, r rune) int {
	 122  	switch {
	 123  	case 0 <= r && r < utf8.RuneSelf:
	 124  		return IndexByte(s, byte(r))
	 125  	case r == utf8.RuneError:
	 126  		for i, r := range s {
	 127  			if r == utf8.RuneError {
	 128  				return i
	 129  			}
	 130  		}
	 131  		return -1
	 132  	case !utf8.ValidRune(r):
	 133  		return -1
	 134  	default:
	 135  		return Index(s, string(r))
	 136  	}
	 137  }
	 138  
	 139  // IndexAny returns the index of the first instance of any Unicode code point
	 140  // from chars in s, or -1 if no Unicode code point from chars is present in s.
	 141  func IndexAny(s, chars string) int {
	 142  	if chars == "" {
	 143  		// Avoid scanning all of s.
	 144  		return -1
	 145  	}
	 146  	if len(chars) == 1 {
	 147  		// Avoid scanning all of s.
	 148  		r := rune(chars[0])
	 149  		if r >= utf8.RuneSelf {
	 150  			r = utf8.RuneError
	 151  		}
	 152  		return IndexRune(s, r)
	 153  	}
	 154  	if len(s) > 8 {
	 155  		if as, isASCII := makeASCIISet(chars); isASCII {
	 156  			for i := 0; i < len(s); i++ {
	 157  				if as.contains(s[i]) {
	 158  					return i
	 159  				}
	 160  			}
	 161  			return -1
	 162  		}
	 163  	}
	 164  	for i, c := range s {
	 165  		if IndexRune(chars, c) >= 0 {
	 166  			return i
	 167  		}
	 168  	}
	 169  	return -1
	 170  }
	 171  
	 172  // LastIndexAny returns the index of the last instance of any Unicode code
	 173  // point from chars in s, or -1 if no Unicode code point from chars is
	 174  // present in s.
	 175  func LastIndexAny(s, chars string) int {
	 176  	if chars == "" {
	 177  		// Avoid scanning all of s.
	 178  		return -1
	 179  	}
	 180  	if len(s) == 1 {
	 181  		rc := rune(s[0])
	 182  		if rc >= utf8.RuneSelf {
	 183  			rc = utf8.RuneError
	 184  		}
	 185  		if IndexRune(chars, rc) >= 0 {
	 186  			return 0
	 187  		}
	 188  		return -1
	 189  	}
	 190  	if len(s) > 8 {
	 191  		if as, isASCII := makeASCIISet(chars); isASCII {
	 192  			for i := len(s) - 1; i >= 0; i-- {
	 193  				if as.contains(s[i]) {
	 194  					return i
	 195  				}
	 196  			}
	 197  			return -1
	 198  		}
	 199  	}
	 200  	if len(chars) == 1 {
	 201  		rc := rune(chars[0])
	 202  		if rc >= utf8.RuneSelf {
	 203  			rc = utf8.RuneError
	 204  		}
	 205  		for i := len(s); i > 0; {
	 206  			r, size := utf8.DecodeLastRuneInString(s[:i])
	 207  			i -= size
	 208  			if rc == r {
	 209  				return i
	 210  			}
	 211  		}
	 212  		return -1
	 213  	}
	 214  	for i := len(s); i > 0; {
	 215  		r, size := utf8.DecodeLastRuneInString(s[:i])
	 216  		i -= size
	 217  		if IndexRune(chars, r) >= 0 {
	 218  			return i
	 219  		}
	 220  	}
	 221  	return -1
	 222  }
	 223  
	 224  // LastIndexByte returns the index of the last instance of c in s, or -1 if c is not present in s.
	 225  func LastIndexByte(s string, c byte) int {
	 226  	for i := len(s) - 1; i >= 0; i-- {
	 227  		if s[i] == c {
	 228  			return i
	 229  		}
	 230  	}
	 231  	return -1
	 232  }
	 233  
	 234  // Generic split: splits after each instance of sep,
	 235  // including sepSave bytes of sep in the subarrays.
	 236  func genSplit(s, sep string, sepSave, n int) []string {
	 237  	if n == 0 {
	 238  		return nil
	 239  	}
	 240  	if sep == "" {
	 241  		return explode(s, n)
	 242  	}
	 243  	if n < 0 {
	 244  		n = Count(s, sep) + 1
	 245  	}
	 246  
	 247  	a := make([]string, n)
	 248  	n--
	 249  	i := 0
	 250  	for i < n {
	 251  		m := Index(s, sep)
	 252  		if m < 0 {
	 253  			break
	 254  		}
	 255  		a[i] = s[:m+sepSave]
	 256  		s = s[m+len(sep):]
	 257  		i++
	 258  	}
	 259  	a[i] = s
	 260  	return a[:i+1]
	 261  }
	 262  
	 263  // SplitN slices s into substrings separated by sep and returns a slice of
	 264  // the substrings between those separators.
	 265  //
	 266  // The count determines the number of substrings to return:
	 267  //	 n > 0: at most n substrings; the last substring will be the unsplit remainder.
	 268  //	 n == 0: the result is nil (zero substrings)
	 269  //	 n < 0: all substrings
	 270  //
	 271  // Edge cases for s and sep (for example, empty strings) are handled
	 272  // as described in the documentation for Split.
	 273  func SplitN(s, sep string, n int) []string { return genSplit(s, sep, 0, n) }
	 274  
	 275  // SplitAfterN slices s into substrings after each instance of sep and
	 276  // returns a slice of those substrings.
	 277  //
	 278  // The count determines the number of substrings to return:
	 279  //	 n > 0: at most n substrings; the last substring will be the unsplit remainder.
	 280  //	 n == 0: the result is nil (zero substrings)
	 281  //	 n < 0: all substrings
	 282  //
	 283  // Edge cases for s and sep (for example, empty strings) are handled
	 284  // as described in the documentation for SplitAfter.
	 285  func SplitAfterN(s, sep string, n int) []string {
	 286  	return genSplit(s, sep, len(sep), n)
	 287  }
	 288  
	 289  // Split slices s into all substrings separated by sep and returns a slice of
	 290  // the substrings between those separators.
	 291  //
	 292  // If s does not contain sep and sep is not empty, Split returns a
	 293  // slice of length 1 whose only element is s.
	 294  //
	 295  // If sep is empty, Split splits after each UTF-8 sequence. If both s
	 296  // and sep are empty, Split returns an empty slice.
	 297  //
	 298  // It is equivalent to SplitN with a count of -1.
	 299  func Split(s, sep string) []string { return genSplit(s, sep, 0, -1) }
	 300  
	 301  // SplitAfter slices s into all substrings after each instance of sep and
	 302  // returns a slice of those substrings.
	 303  //
	 304  // If s does not contain sep and sep is not empty, SplitAfter returns
	 305  // a slice of length 1 whose only element is s.
	 306  //
	 307  // If sep is empty, SplitAfter splits after each UTF-8 sequence. If
	 308  // both s and sep are empty, SplitAfter returns an empty slice.
	 309  //
	 310  // It is equivalent to SplitAfterN with a count of -1.
	 311  func SplitAfter(s, sep string) []string {
	 312  	return genSplit(s, sep, len(sep), -1)
	 313  }
	 314  
	 315  var asciiSpace = [256]uint8{'\t': 1, '\n': 1, '\v': 1, '\f': 1, '\r': 1, ' ': 1}
	 316  
	 317  // Fields splits the string s around each instance of one or more consecutive white space
	 318  // characters, as defined by unicode.IsSpace, returning a slice of substrings of s or an
	 319  // empty slice if s contains only white space.
	 320  func Fields(s string) []string {
	 321  	// First count the fields.
	 322  	// This is an exact count if s is ASCII, otherwise it is an approximation.
	 323  	n := 0
	 324  	wasSpace := 1
	 325  	// setBits is used to track which bits are set in the bytes of s.
	 326  	setBits := uint8(0)
	 327  	for i := 0; i < len(s); i++ {
	 328  		r := s[i]
	 329  		setBits |= r
	 330  		isSpace := int(asciiSpace[r])
	 331  		n += wasSpace & ^isSpace
	 332  		wasSpace = isSpace
	 333  	}
	 334  
	 335  	if setBits >= utf8.RuneSelf {
	 336  		// Some runes in the input string are not ASCII.
	 337  		return FieldsFunc(s, unicode.IsSpace)
	 338  	}
	 339  	// ASCII fast path
	 340  	a := make([]string, n)
	 341  	na := 0
	 342  	fieldStart := 0
	 343  	i := 0
	 344  	// Skip spaces in the front of the input.
	 345  	for i < len(s) && asciiSpace[s[i]] != 0 {
	 346  		i++
	 347  	}
	 348  	fieldStart = i
	 349  	for i < len(s) {
	 350  		if asciiSpace[s[i]] == 0 {
	 351  			i++
	 352  			continue
	 353  		}
	 354  		a[na] = s[fieldStart:i]
	 355  		na++
	 356  		i++
	 357  		// Skip spaces in between fields.
	 358  		for i < len(s) && asciiSpace[s[i]] != 0 {
	 359  			i++
	 360  		}
	 361  		fieldStart = i
	 362  	}
	 363  	if fieldStart < len(s) { // Last field might end at EOF.
	 364  		a[na] = s[fieldStart:]
	 365  	}
	 366  	return a
	 367  }
	 368  
	 369  // FieldsFunc splits the string s at each run of Unicode code points c satisfying f(c)
	 370  // and returns an array of slices of s. If all code points in s satisfy f(c) or the
	 371  // string is empty, an empty slice is returned.
	 372  //
	 373  // FieldsFunc makes no guarantees about the order in which it calls f(c)
	 374  // and assumes that f always returns the same value for a given c.
	 375  func FieldsFunc(s string, f func(rune) bool) []string {
	 376  	// A span is used to record a slice of s of the form s[start:end].
	 377  	// The start index is inclusive and the end index is exclusive.
	 378  	type span struct {
	 379  		start int
	 380  		end	 int
	 381  	}
	 382  	spans := make([]span, 0, 32)
	 383  
	 384  	// Find the field start and end indices.
	 385  	// Doing this in a separate pass (rather than slicing the string s
	 386  	// and collecting the result substrings right away) is significantly
	 387  	// more efficient, possibly due to cache effects.
	 388  	start := -1 // valid span start if >= 0
	 389  	for end, rune := range s {
	 390  		if f(rune) {
	 391  			if start >= 0 {
	 392  				spans = append(spans, span{start, end})
	 393  				// Set start to a negative value.
	 394  				// Note: using -1 here consistently and reproducibly
	 395  				// slows down this code by a several percent on amd64.
	 396  				start = ^start
	 397  			}
	 398  		} else {
	 399  			if start < 0 {
	 400  				start = end
	 401  			}
	 402  		}
	 403  	}
	 404  
	 405  	// Last field might end at EOF.
	 406  	if start >= 0 {
	 407  		spans = append(spans, span{start, len(s)})
	 408  	}
	 409  
	 410  	// Create strings from recorded field indices.
	 411  	a := make([]string, len(spans))
	 412  	for i, span := range spans {
	 413  		a[i] = s[span.start:span.end]
	 414  	}
	 415  
	 416  	return a
	 417  }
	 418  
	 419  // Join concatenates the elements of its first argument to create a single string. The separator
	 420  // string sep is placed between elements in the resulting string.
	 421  func Join(elems []string, sep string) string {
	 422  	switch len(elems) {
	 423  	case 0:
	 424  		return ""
	 425  	case 1:
	 426  		return elems[0]
	 427  	}
	 428  	n := len(sep) * (len(elems) - 1)
	 429  	for i := 0; i < len(elems); i++ {
	 430  		n += len(elems[i])
	 431  	}
	 432  
	 433  	var b Builder
	 434  	b.Grow(n)
	 435  	b.WriteString(elems[0])
	 436  	for _, s := range elems[1:] {
	 437  		b.WriteString(sep)
	 438  		b.WriteString(s)
	 439  	}
	 440  	return b.String()
	 441  }
	 442  
	 443  // HasPrefix tests whether the string s begins with prefix.
	 444  func HasPrefix(s, prefix string) bool {
	 445  	return len(s) >= len(prefix) && s[0:len(prefix)] == prefix
	 446  }
	 447  
	 448  // HasSuffix tests whether the string s ends with suffix.
	 449  func HasSuffix(s, suffix string) bool {
	 450  	return len(s) >= len(suffix) && s[len(s)-len(suffix):] == suffix
	 451  }
	 452  
	 453  // Map returns a copy of the string s with all its characters modified
	 454  // according to the mapping function. If mapping returns a negative value, the character is
	 455  // dropped from the string with no replacement.
	 456  func Map(mapping func(rune) rune, s string) string {
	 457  	// In the worst case, the string can grow when mapped, making
	 458  	// things unpleasant. But it's so rare we barge in assuming it's
	 459  	// fine. It could also shrink but that falls out naturally.
	 460  
	 461  	// The output buffer b is initialized on demand, the first
	 462  	// time a character differs.
	 463  	var b Builder
	 464  
	 465  	for i, c := range s {
	 466  		r := mapping(c)
	 467  		if r == c && c != utf8.RuneError {
	 468  			continue
	 469  		}
	 470  
	 471  		var width int
	 472  		if c == utf8.RuneError {
	 473  			c, width = utf8.DecodeRuneInString(s[i:])
	 474  			if width != 1 && r == c {
	 475  				continue
	 476  			}
	 477  		} else {
	 478  			width = utf8.RuneLen(c)
	 479  		}
	 480  
	 481  		b.Grow(len(s) + utf8.UTFMax)
	 482  		b.WriteString(s[:i])
	 483  		if r >= 0 {
	 484  			b.WriteRune(r)
	 485  		}
	 486  
	 487  		s = s[i+width:]
	 488  		break
	 489  	}
	 490  
	 491  	// Fast path for unchanged input
	 492  	if b.Cap() == 0 { // didn't call b.Grow above
	 493  		return s
	 494  	}
	 495  
	 496  	for _, c := range s {
	 497  		r := mapping(c)
	 498  
	 499  		if r >= 0 {
	 500  			// common case
	 501  			// Due to inlining, it is more performant to determine if WriteByte should be
	 502  			// invoked rather than always call WriteRune
	 503  			if r < utf8.RuneSelf {
	 504  				b.WriteByte(byte(r))
	 505  			} else {
	 506  				// r is not a ASCII rune.
	 507  				b.WriteRune(r)
	 508  			}
	 509  		}
	 510  	}
	 511  
	 512  	return b.String()
	 513  }
	 514  
	 515  // Repeat returns a new string consisting of count copies of the string s.
	 516  //
	 517  // It panics if count is negative or if
	 518  // the result of (len(s) * count) overflows.
	 519  func Repeat(s string, count int) string {
	 520  	if count == 0 {
	 521  		return ""
	 522  	}
	 523  
	 524  	// Since we cannot return an error on overflow,
	 525  	// we should panic if the repeat will generate
	 526  	// an overflow.
	 527  	// See Issue golang.org/issue/16237
	 528  	if count < 0 {
	 529  		panic("strings: negative Repeat count")
	 530  	} else if len(s)*count/count != len(s) {
	 531  		panic("strings: Repeat count causes overflow")
	 532  	}
	 533  
	 534  	n := len(s) * count
	 535  	var b Builder
	 536  	b.Grow(n)
	 537  	b.WriteString(s)
	 538  	for b.Len() < n {
	 539  		if b.Len() <= n/2 {
	 540  			b.WriteString(b.String())
	 541  		} else {
	 542  			b.WriteString(b.String()[:n-b.Len()])
	 543  			break
	 544  		}
	 545  	}
	 546  	return b.String()
	 547  }
	 548  
	 549  // ToUpper returns s with all Unicode letters mapped to their upper case.
	 550  func ToUpper(s string) string {
	 551  	isASCII, hasLower := true, false
	 552  	for i := 0; i < len(s); i++ {
	 553  		c := s[i]
	 554  		if c >= utf8.RuneSelf {
	 555  			isASCII = false
	 556  			break
	 557  		}
	 558  		hasLower = hasLower || ('a' <= c && c <= 'z')
	 559  	}
	 560  
	 561  	if isASCII { // optimize for ASCII-only strings.
	 562  		if !hasLower {
	 563  			return s
	 564  		}
	 565  		var b Builder
	 566  		b.Grow(len(s))
	 567  		for i := 0; i < len(s); i++ {
	 568  			c := s[i]
	 569  			if 'a' <= c && c <= 'z' {
	 570  				c -= 'a' - 'A'
	 571  			}
	 572  			b.WriteByte(c)
	 573  		}
	 574  		return b.String()
	 575  	}
	 576  	return Map(unicode.ToUpper, s)
	 577  }
	 578  
	 579  // ToLower returns s with all Unicode letters mapped to their lower case.
	 580  func ToLower(s string) string {
	 581  	isASCII, hasUpper := true, false
	 582  	for i := 0; i < len(s); i++ {
	 583  		c := s[i]
	 584  		if c >= utf8.RuneSelf {
	 585  			isASCII = false
	 586  			break
	 587  		}
	 588  		hasUpper = hasUpper || ('A' <= c && c <= 'Z')
	 589  	}
	 590  
	 591  	if isASCII { // optimize for ASCII-only strings.
	 592  		if !hasUpper {
	 593  			return s
	 594  		}
	 595  		var b Builder
	 596  		b.Grow(len(s))
	 597  		for i := 0; i < len(s); i++ {
	 598  			c := s[i]
	 599  			if 'A' <= c && c <= 'Z' {
	 600  				c += 'a' - 'A'
	 601  			}
	 602  			b.WriteByte(c)
	 603  		}
	 604  		return b.String()
	 605  	}
	 606  	return Map(unicode.ToLower, s)
	 607  }
	 608  
	 609  // ToTitle returns a copy of the string s with all Unicode letters mapped to
	 610  // their Unicode title case.
	 611  func ToTitle(s string) string { return Map(unicode.ToTitle, s) }
	 612  
	 613  // ToUpperSpecial returns a copy of the string s with all Unicode letters mapped to their
	 614  // upper case using the case mapping specified by c.
	 615  func ToUpperSpecial(c unicode.SpecialCase, s string) string {
	 616  	return Map(c.ToUpper, s)
	 617  }
	 618  
	 619  // ToLowerSpecial returns a copy of the string s with all Unicode letters mapped to their
	 620  // lower case using the case mapping specified by c.
	 621  func ToLowerSpecial(c unicode.SpecialCase, s string) string {
	 622  	return Map(c.ToLower, s)
	 623  }
	 624  
	 625  // ToTitleSpecial returns a copy of the string s with all Unicode letters mapped to their
	 626  // Unicode title case, giving priority to the special casing rules.
	 627  func ToTitleSpecial(c unicode.SpecialCase, s string) string {
	 628  	return Map(c.ToTitle, s)
	 629  }
	 630  
	 631  // ToValidUTF8 returns a copy of the string s with each run of invalid UTF-8 byte sequences
	 632  // replaced by the replacement string, which may be empty.
	 633  func ToValidUTF8(s, replacement string) string {
	 634  	var b Builder
	 635  
	 636  	for i, c := range s {
	 637  		if c != utf8.RuneError {
	 638  			continue
	 639  		}
	 640  
	 641  		_, wid := utf8.DecodeRuneInString(s[i:])
	 642  		if wid == 1 {
	 643  			b.Grow(len(s) + len(replacement))
	 644  			b.WriteString(s[:i])
	 645  			s = s[i:]
	 646  			break
	 647  		}
	 648  	}
	 649  
	 650  	// Fast path for unchanged input
	 651  	if b.Cap() == 0 { // didn't call b.Grow above
	 652  		return s
	 653  	}
	 654  
	 655  	invalid := false // previous byte was from an invalid UTF-8 sequence
	 656  	for i := 0; i < len(s); {
	 657  		c := s[i]
	 658  		if c < utf8.RuneSelf {
	 659  			i++
	 660  			invalid = false
	 661  			b.WriteByte(c)
	 662  			continue
	 663  		}
	 664  		_, wid := utf8.DecodeRuneInString(s[i:])
	 665  		if wid == 1 {
	 666  			i++
	 667  			if !invalid {
	 668  				invalid = true
	 669  				b.WriteString(replacement)
	 670  			}
	 671  			continue
	 672  		}
	 673  		invalid = false
	 674  		b.WriteString(s[i : i+wid])
	 675  		i += wid
	 676  	}
	 677  
	 678  	return b.String()
	 679  }
	 680  
	 681  // isSeparator reports whether the rune could mark a word boundary.
	 682  // TODO: update when package unicode captures more of the properties.
	 683  func isSeparator(r rune) bool {
	 684  	// ASCII alphanumerics and underscore are not separators
	 685  	if r <= 0x7F {
	 686  		switch {
	 687  		case '0' <= r && r <= '9':
	 688  			return false
	 689  		case 'a' <= r && r <= 'z':
	 690  			return false
	 691  		case 'A' <= r && r <= 'Z':
	 692  			return false
	 693  		case r == '_':
	 694  			return false
	 695  		}
	 696  		return true
	 697  	}
	 698  	// Letters and digits are not separators
	 699  	if unicode.IsLetter(r) || unicode.IsDigit(r) {
	 700  		return false
	 701  	}
	 702  	// Otherwise, all we can do for now is treat spaces as separators.
	 703  	return unicode.IsSpace(r)
	 704  }
	 705  
	 706  // Title returns a copy of the string s with all Unicode letters that begin words
	 707  // mapped to their Unicode title case.
	 708  //
	 709  // BUG(rsc): The rule Title uses for word boundaries does not handle Unicode punctuation properly.
	 710  func Title(s string) string {
	 711  	// Use a closure here to remember state.
	 712  	// Hackish but effective. Depends on Map scanning in order and calling
	 713  	// the closure once per rune.
	 714  	prev := ' '
	 715  	return Map(
	 716  		func(r rune) rune {
	 717  			if isSeparator(prev) {
	 718  				prev = r
	 719  				return unicode.ToTitle(r)
	 720  			}
	 721  			prev = r
	 722  			return r
	 723  		},
	 724  		s)
	 725  }
	 726  
	 727  // TrimLeftFunc returns a slice of the string s with all leading
	 728  // Unicode code points c satisfying f(c) removed.
	 729  func TrimLeftFunc(s string, f func(rune) bool) string {
	 730  	i := indexFunc(s, f, false)
	 731  	if i == -1 {
	 732  		return ""
	 733  	}
	 734  	return s[i:]
	 735  }
	 736  
	 737  // TrimRightFunc returns a slice of the string s with all trailing
	 738  // Unicode code points c satisfying f(c) removed.
	 739  func TrimRightFunc(s string, f func(rune) bool) string {
	 740  	i := lastIndexFunc(s, f, false)
	 741  	if i >= 0 && s[i] >= utf8.RuneSelf {
	 742  		_, wid := utf8.DecodeRuneInString(s[i:])
	 743  		i += wid
	 744  	} else {
	 745  		i++
	 746  	}
	 747  	return s[0:i]
	 748  }
	 749  
	 750  // TrimFunc returns a slice of the string s with all leading
	 751  // and trailing Unicode code points c satisfying f(c) removed.
	 752  func TrimFunc(s string, f func(rune) bool) string {
	 753  	return TrimRightFunc(TrimLeftFunc(s, f), f)
	 754  }
	 755  
	 756  // IndexFunc returns the index into s of the first Unicode
	 757  // code point satisfying f(c), or -1 if none do.
	 758  func IndexFunc(s string, f func(rune) bool) int {
	 759  	return indexFunc(s, f, true)
	 760  }
	 761  
	 762  // LastIndexFunc returns the index into s of the last
	 763  // Unicode code point satisfying f(c), or -1 if none do.
	 764  func LastIndexFunc(s string, f func(rune) bool) int {
	 765  	return lastIndexFunc(s, f, true)
	 766  }
	 767  
	 768  // indexFunc is the same as IndexFunc except that if
	 769  // truth==false, the sense of the predicate function is
	 770  // inverted.
	 771  func indexFunc(s string, f func(rune) bool, truth bool) int {
	 772  	for i, r := range s {
	 773  		if f(r) == truth {
	 774  			return i
	 775  		}
	 776  	}
	 777  	return -1
	 778  }
	 779  
	 780  // lastIndexFunc is the same as LastIndexFunc except that if
	 781  // truth==false, the sense of the predicate function is
	 782  // inverted.
	 783  func lastIndexFunc(s string, f func(rune) bool, truth bool) int {
	 784  	for i := len(s); i > 0; {
	 785  		r, size := utf8.DecodeLastRuneInString(s[0:i])
	 786  		i -= size
	 787  		if f(r) == truth {
	 788  			return i
	 789  		}
	 790  	}
	 791  	return -1
	 792  }
	 793  
	 794  // asciiSet is a 32-byte value, where each bit represents the presence of a
	 795  // given ASCII character in the set. The 128-bits of the lower 16 bytes,
	 796  // starting with the least-significant bit of the lowest word to the
	 797  // most-significant bit of the highest word, map to the full range of all
	 798  // 128 ASCII characters. The 128-bits of the upper 16 bytes will be zeroed,
	 799  // ensuring that any non-ASCII character will be reported as not in the set.
	 800  type asciiSet [8]uint32
	 801  
	 802  // makeASCIISet creates a set of ASCII characters and reports whether all
	 803  // characters in chars are ASCII.
	 804  func makeASCIISet(chars string) (as asciiSet, ok bool) {
	 805  	for i := 0; i < len(chars); i++ {
	 806  		c := chars[i]
	 807  		if c >= utf8.RuneSelf {
	 808  			return as, false
	 809  		}
	 810  		as[c>>5] |= 1 << uint(c&31)
	 811  	}
	 812  	return as, true
	 813  }
	 814  
	 815  // contains reports whether c is inside the set.
	 816  func (as *asciiSet) contains(c byte) bool {
	 817  	return (as[c>>5] & (1 << uint(c&31))) != 0
	 818  }
	 819  
	 820  func makeCutsetFunc(cutset string) func(rune) bool {
	 821  	if len(cutset) == 1 && cutset[0] < utf8.RuneSelf {
	 822  		return func(r rune) bool {
	 823  			return r == rune(cutset[0])
	 824  		}
	 825  	}
	 826  	if as, isASCII := makeASCIISet(cutset); isASCII {
	 827  		return func(r rune) bool {
	 828  			return r < utf8.RuneSelf && as.contains(byte(r))
	 829  		}
	 830  	}
	 831  	return func(r rune) bool { return IndexRune(cutset, r) >= 0 }
	 832  }
	 833  
	 834  // Trim returns a slice of the string s with all leading and
	 835  // trailing Unicode code points contained in cutset removed.
	 836  func Trim(s, cutset string) string {
	 837  	if s == "" || cutset == "" {
	 838  		return s
	 839  	}
	 840  	return TrimFunc(s, makeCutsetFunc(cutset))
	 841  }
	 842  
	 843  // TrimLeft returns a slice of the string s with all leading
	 844  // Unicode code points contained in cutset removed.
	 845  //
	 846  // To remove a prefix, use TrimPrefix instead.
	 847  func TrimLeft(s, cutset string) string {
	 848  	if s == "" || cutset == "" {
	 849  		return s
	 850  	}
	 851  	return TrimLeftFunc(s, makeCutsetFunc(cutset))
	 852  }
	 853  
	 854  // TrimRight returns a slice of the string s, with all trailing
	 855  // Unicode code points contained in cutset removed.
	 856  //
	 857  // To remove a suffix, use TrimSuffix instead.
	 858  func TrimRight(s, cutset string) string {
	 859  	if s == "" || cutset == "" {
	 860  		return s
	 861  	}
	 862  	return TrimRightFunc(s, makeCutsetFunc(cutset))
	 863  }
	 864  
	 865  // TrimSpace returns a slice of the string s, with all leading
	 866  // and trailing white space removed, as defined by Unicode.
	 867  func TrimSpace(s string) string {
	 868  	// Fast path for ASCII: look for the first ASCII non-space byte
	 869  	start := 0
	 870  	for ; start < len(s); start++ {
	 871  		c := s[start]
	 872  		if c >= utf8.RuneSelf {
	 873  			// If we run into a non-ASCII byte, fall back to the
	 874  			// slower unicode-aware method on the remaining bytes
	 875  			return TrimFunc(s[start:], unicode.IsSpace)
	 876  		}
	 877  		if asciiSpace[c] == 0 {
	 878  			break
	 879  		}
	 880  	}
	 881  
	 882  	// Now look for the first ASCII non-space byte from the end
	 883  	stop := len(s)
	 884  	for ; stop > start; stop-- {
	 885  		c := s[stop-1]
	 886  		if c >= utf8.RuneSelf {
	 887  			return TrimFunc(s[start:stop], unicode.IsSpace)
	 888  		}
	 889  		if asciiSpace[c] == 0 {
	 890  			break
	 891  		}
	 892  	}
	 893  
	 894  	// At this point s[start:stop] starts and ends with an ASCII
	 895  	// non-space bytes, so we're done. Non-ASCII cases have already
	 896  	// been handled above.
	 897  	return s[start:stop]
	 898  }
	 899  
	 900  // TrimPrefix returns s without the provided leading prefix string.
	 901  // If s doesn't start with prefix, s is returned unchanged.
	 902  func TrimPrefix(s, prefix string) string {
	 903  	if HasPrefix(s, prefix) {
	 904  		return s[len(prefix):]
	 905  	}
	 906  	return s
	 907  }
	 908  
	 909  // TrimSuffix returns s without the provided trailing suffix string.
	 910  // If s doesn't end with suffix, s is returned unchanged.
	 911  func TrimSuffix(s, suffix string) string {
	 912  	if HasSuffix(s, suffix) {
	 913  		return s[:len(s)-len(suffix)]
	 914  	}
	 915  	return s
	 916  }
	 917  
	 918  // Replace returns a copy of the string s with the first n
	 919  // non-overlapping instances of old replaced by new.
	 920  // If old is empty, it matches at the beginning of the string
	 921  // and after each UTF-8 sequence, yielding up to k+1 replacements
	 922  // for a k-rune string.
	 923  // If n < 0, there is no limit on the number of replacements.
	 924  func Replace(s, old, new string, n int) string {
	 925  	if old == new || n == 0 {
	 926  		return s // avoid allocation
	 927  	}
	 928  
	 929  	// Compute number of replacements.
	 930  	if m := Count(s, old); m == 0 {
	 931  		return s // avoid allocation
	 932  	} else if n < 0 || m < n {
	 933  		n = m
	 934  	}
	 935  
	 936  	// Apply replacements to buffer.
	 937  	var b Builder
	 938  	b.Grow(len(s) + n*(len(new)-len(old)))
	 939  	start := 0
	 940  	for i := 0; i < n; i++ {
	 941  		j := start
	 942  		if len(old) == 0 {
	 943  			if i > 0 {
	 944  				_, wid := utf8.DecodeRuneInString(s[start:])
	 945  				j += wid
	 946  			}
	 947  		} else {
	 948  			j += Index(s[start:], old)
	 949  		}
	 950  		b.WriteString(s[start:j])
	 951  		b.WriteString(new)
	 952  		start = j + len(old)
	 953  	}
	 954  	b.WriteString(s[start:])
	 955  	return b.String()
	 956  }
	 957  
	 958  // ReplaceAll returns a copy of the string s with all
	 959  // non-overlapping instances of old replaced by new.
	 960  // If old is empty, it matches at the beginning of the string
	 961  // and after each UTF-8 sequence, yielding up to k+1 replacements
	 962  // for a k-rune string.
	 963  func ReplaceAll(s, old, new string) string {
	 964  	return Replace(s, old, new, -1)
	 965  }
	 966  
	 967  // EqualFold reports whether s and t, interpreted as UTF-8 strings,
	 968  // are equal under Unicode case-folding, which is a more general
	 969  // form of case-insensitivity.
	 970  func EqualFold(s, t string) bool {
	 971  	for s != "" && t != "" {
	 972  		// Extract first rune from each string.
	 973  		var sr, tr rune
	 974  		if s[0] < utf8.RuneSelf {
	 975  			sr, s = rune(s[0]), s[1:]
	 976  		} else {
	 977  			r, size := utf8.DecodeRuneInString(s)
	 978  			sr, s = r, s[size:]
	 979  		}
	 980  		if t[0] < utf8.RuneSelf {
	 981  			tr, t = rune(t[0]), t[1:]
	 982  		} else {
	 983  			r, size := utf8.DecodeRuneInString(t)
	 984  			tr, t = r, t[size:]
	 985  		}
	 986  
	 987  		// If they match, keep going; if not, return false.
	 988  
	 989  		// Easy case.
	 990  		if tr == sr {
	 991  			continue
	 992  		}
	 993  
	 994  		// Make sr < tr to simplify what follows.
	 995  		if tr < sr {
	 996  			tr, sr = sr, tr
	 997  		}
	 998  		// Fast check for ASCII.
	 999  		if tr < utf8.RuneSelf {
	1000  			// ASCII only, sr/tr must be upper/lower case
	1001  			if 'A' <= sr && sr <= 'Z' && tr == sr+'a'-'A' {
	1002  				continue
	1003  			}
	1004  			return false
	1005  		}
	1006  
	1007  		// General case. SimpleFold(x) returns the next equivalent rune > x
	1008  		// or wraps around to smaller values.
	1009  		r := unicode.SimpleFold(sr)
	1010  		for r != sr && r < tr {
	1011  			r = unicode.SimpleFold(r)
	1012  		}
	1013  		if r == tr {
	1014  			continue
	1015  		}
	1016  		return false
	1017  	}
	1018  
	1019  	// One string is empty. Are both?
	1020  	return s == t
	1021  }
	1022  
	1023  // Index returns the index of the first instance of substr in s, or -1 if substr is not present in s.
	1024  func Index(s, substr string) int {
	1025  	n := len(substr)
	1026  	switch {
	1027  	case n == 0:
	1028  		return 0
	1029  	case n == 1:
	1030  		return IndexByte(s, substr[0])
	1031  	case n == len(s):
	1032  		if substr == s {
	1033  			return 0
	1034  		}
	1035  		return -1
	1036  	case n > len(s):
	1037  		return -1
	1038  	case n <= bytealg.MaxLen:
	1039  		// Use brute force when s and substr both are small
	1040  		if len(s) <= bytealg.MaxBruteForce {
	1041  			return bytealg.IndexString(s, substr)
	1042  		}
	1043  		c0 := substr[0]
	1044  		c1 := substr[1]
	1045  		i := 0
	1046  		t := len(s) - n + 1
	1047  		fails := 0
	1048  		for i < t {
	1049  			if s[i] != c0 {
	1050  				// IndexByte is faster than bytealg.IndexString, so use it as long as
	1051  				// we're not getting lots of false positives.
	1052  				o := IndexByte(s[i+1:t], c0)
	1053  				if o < 0 {
	1054  					return -1
	1055  				}
	1056  				i += o + 1
	1057  			}
	1058  			if s[i+1] == c1 && s[i:i+n] == substr {
	1059  				return i
	1060  			}
	1061  			fails++
	1062  			i++
	1063  			// Switch to bytealg.IndexString when IndexByte produces too many false positives.
	1064  			if fails > bytealg.Cutover(i) {
	1065  				r := bytealg.IndexString(s[i:], substr)
	1066  				if r >= 0 {
	1067  					return r + i
	1068  				}
	1069  				return -1
	1070  			}
	1071  		}
	1072  		return -1
	1073  	}
	1074  	c0 := substr[0]
	1075  	c1 := substr[1]
	1076  	i := 0
	1077  	t := len(s) - n + 1
	1078  	fails := 0
	1079  	for i < t {
	1080  		if s[i] != c0 {
	1081  			o := IndexByte(s[i+1:t], c0)
	1082  			if o < 0 {
	1083  				return -1
	1084  			}
	1085  			i += o + 1
	1086  		}
	1087  		if s[i+1] == c1 && s[i:i+n] == substr {
	1088  			return i
	1089  		}
	1090  		i++
	1091  		fails++
	1092  		if fails >= 4+i>>4 && i < t {
	1093  			// See comment in ../bytes/bytes.go.
	1094  			j := bytealg.IndexRabinKarp(s[i:], substr)
	1095  			if j < 0 {
	1096  				return -1
	1097  			}
	1098  			return i + j
	1099  		}
	1100  	}
	1101  	return -1
	1102  }
	1103  

View as plain text