Source file src/bytes/bytes_test.go

     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_test
     6  
     7  import (
     8  	. "bytes"
     9  	"fmt"
    10  	"internal/asan"
    11  	"internal/testenv"
    12  	"iter"
    13  	"math"
    14  	"math/rand"
    15  	"slices"
    16  	"strings"
    17  	"testing"
    18  	"unicode"
    19  	"unicode/utf8"
    20  	"unsafe"
    21  )
    22  
    23  func sliceOfString(s [][]byte) []string {
    24  	result := make([]string, len(s))
    25  	for i, v := range s {
    26  		result[i] = string(v)
    27  	}
    28  	return result
    29  }
    30  
    31  func collect(t *testing.T, seq iter.Seq[[]byte]) [][]byte {
    32  	out := slices.Collect(seq)
    33  	out1 := slices.Collect(seq)
    34  	if !slices.Equal(sliceOfString(out), sliceOfString(out1)) {
    35  		t.Fatalf("inconsistent seq:\n%s\n%s", out, out1)
    36  	}
    37  	return out
    38  }
    39  
    40  type LinesTest struct {
    41  	a string
    42  	b []string
    43  }
    44  
    45  var linesTests = []LinesTest{
    46  	{a: "abc\nabc\n", b: []string{"abc\n", "abc\n"}},
    47  	{a: "abc\r\nabc", b: []string{"abc\r\n", "abc"}},
    48  	{a: "abc\r\n", b: []string{"abc\r\n"}},
    49  	{a: "\nabc", b: []string{"\n", "abc"}},
    50  	{a: "\nabc\n\n", b: []string{"\n", "abc\n", "\n"}},
    51  }
    52  
    53  func TestLines(t *testing.T) {
    54  	for _, s := range linesTests {
    55  		result := sliceOfString(slices.Collect(Lines([]byte(s.a))))
    56  		if !slices.Equal(result, s.b) {
    57  			t.Errorf(`slices.Collect(Lines(%q)) = %q; want %q`, s.a, result, s.b)
    58  		}
    59  	}
    60  }
    61  
    62  // For ease of reading, the test cases use strings that are converted to byte
    63  // slices before invoking the functions.
    64  
    65  var abcd = "abcd"
    66  var faces = "☺☻☹"
    67  var commas = "1,2,3,4"
    68  var dots = "1....2....3....4"
    69  
    70  type BinOpTest struct {
    71  	a string
    72  	b string
    73  	i int
    74  }
    75  
    76  func TestEqual(t *testing.T) {
    77  	// Run the tests and check for allocation at the same time.
    78  	allocs := testing.AllocsPerRun(10, func() {
    79  		for _, tt := range compareTests {
    80  			eql := Equal(tt.a, tt.b)
    81  			if eql != (tt.i == 0) {
    82  				t.Errorf(`Equal(%q, %q) = %v`, tt.a, tt.b, eql)
    83  			}
    84  		}
    85  	})
    86  	if allocs > 0 {
    87  		t.Errorf("Equal allocated %v times", allocs)
    88  	}
    89  }
    90  
    91  func TestEqualExhaustive(t *testing.T) {
    92  	var size = 128
    93  	if testing.Short() {
    94  		size = 32
    95  	}
    96  	a := make([]byte, size)
    97  	b := make([]byte, size)
    98  	b_init := make([]byte, size)
    99  	// randomish but deterministic data
   100  	for i := 0; i < size; i++ {
   101  		a[i] = byte(17 * i)
   102  		b_init[i] = byte(23*i + 100)
   103  	}
   104  
   105  	for len := 0; len <= size; len++ {
   106  		for x := 0; x <= size-len; x++ {
   107  			for y := 0; y <= size-len; y++ {
   108  				copy(b, b_init)
   109  				copy(b[y:y+len], a[x:x+len])
   110  				if !Equal(a[x:x+len], b[y:y+len]) || !Equal(b[y:y+len], a[x:x+len]) {
   111  					t.Errorf("Equal(%d, %d, %d) = false", len, x, y)
   112  				}
   113  			}
   114  		}
   115  	}
   116  }
   117  
   118  // make sure Equal returns false for minimally different strings. The data
   119  // is all zeros except for a single one in one location.
   120  func TestNotEqual(t *testing.T) {
   121  	var size = 128
   122  	if testing.Short() {
   123  		size = 32
   124  	}
   125  	a := make([]byte, size)
   126  	b := make([]byte, size)
   127  
   128  	for len := 0; len <= size; len++ {
   129  		for x := 0; x <= size-len; x++ {
   130  			for y := 0; y <= size-len; y++ {
   131  				for diffpos := x; diffpos < x+len; diffpos++ {
   132  					a[diffpos] = 1
   133  					if Equal(a[x:x+len], b[y:y+len]) || Equal(b[y:y+len], a[x:x+len]) {
   134  						t.Errorf("NotEqual(%d, %d, %d, %d) = true", len, x, y, diffpos)
   135  					}
   136  					a[diffpos] = 0
   137  				}
   138  			}
   139  		}
   140  	}
   141  }
   142  
   143  var indexTests = []BinOpTest{
   144  	{"", "", 0},
   145  	{"", "a", -1},
   146  	{"", "foo", -1},
   147  	{"fo", "foo", -1},
   148  	{"foo", "baz", -1},
   149  	{"foo", "foo", 0},
   150  	{"oofofoofooo", "f", 2},
   151  	{"oofofoofooo", "foo", 4},
   152  	{"barfoobarfoo", "foo", 3},
   153  	{"foo", "", 0},
   154  	{"foo", "o", 1},
   155  	{"abcABCabc", "A", 3},
   156  	// cases with one byte strings - test IndexByte and special case in Index()
   157  	{"", "a", -1},
   158  	{"x", "a", -1},
   159  	{"x", "x", 0},
   160  	{"abc", "a", 0},
   161  	{"abc", "b", 1},
   162  	{"abc", "c", 2},
   163  	{"abc", "x", -1},
   164  	{"barfoobarfooyyyzzzyyyzzzyyyzzzyyyxxxzzzyyy", "x", 33},
   165  	{"fofofofooofoboo", "oo", 7},
   166  	{"fofofofofofoboo", "ob", 11},
   167  	{"fofofofofofoboo", "boo", 12},
   168  	{"fofofofofofoboo", "oboo", 11},
   169  	{"fofofofofoooboo", "fooo", 8},
   170  	{"fofofofofofoboo", "foboo", 10},
   171  	{"fofofofofofoboo", "fofob", 8},
   172  	{"fofofofofofofoffofoobarfoo", "foffof", 12},
   173  	{"fofofofofoofofoffofoobarfoo", "foffof", 13},
   174  	{"fofofofofofofoffofoobarfoo", "foffofo", 12},
   175  	{"fofofofofoofofoffofoobarfoo", "foffofo", 13},
   176  	{"fofofofofoofofoffofoobarfoo", "foffofoo", 13},
   177  	{"fofofofofofofoffofoobarfoo", "foffofoo", 12},
   178  	{"fofofofofoofofoffofoobarfoo", "foffofoob", 13},
   179  	{"fofofofofofofoffofoobarfoo", "foffofoob", 12},
   180  	{"fofofofofoofofoffofoobarfoo", "foffofooba", 13},
   181  	{"fofofofofofofoffofoobarfoo", "foffofooba", 12},
   182  	{"fofofofofoofofoffofoobarfoo", "foffofoobar", 13},
   183  	{"fofofofofofofoffofoobarfoo", "foffofoobar", 12},
   184  	{"fofofofofoofofoffofoobarfoo", "foffofoobarf", 13},
   185  	{"fofofofofofofoffofoobarfoo", "foffofoobarf", 12},
   186  	{"fofofofofoofofoffofoobarfoo", "foffofoobarfo", 13},
   187  	{"fofofofofofofoffofoobarfoo", "foffofoobarfo", 12},
   188  	{"fofofofofoofofoffofoobarfoo", "foffofoobarfoo", 13},
   189  	{"fofofofofofofoffofoobarfoo", "foffofoobarfoo", 12},
   190  	{"fofofofofoofofoffofoobarfoo", "ofoffofoobarfoo", 12},
   191  	{"fofofofofofofoffofoobarfoo", "ofoffofoobarfoo", 11},
   192  	{"fofofofofoofofoffofoobarfoo", "fofoffofoobarfoo", 11},
   193  	{"fofofofofofofoffofoobarfoo", "fofoffofoobarfoo", 10},
   194  	{"fofofofofoofofoffofoobarfoo", "foobars", -1},
   195  	{"foofyfoobarfoobar", "y", 4},
   196  	{"oooooooooooooooooooooo", "r", -1},
   197  	{"oxoxoxoxoxoxoxoxoxoxoxoy", "oy", 22},
   198  	{"oxoxoxoxoxoxoxoxoxoxoxox", "oy", -1},
   199  	// test fallback to Rabin-Karp.
   200  	{"000000000000000000000000000000000000000000000000000000000000000000000001", "0000000000000000000000000000000000000000000000000000000000000000001", 5},
   201  	// test fallback to IndexRune
   202  	{"oxoxoxoxoxoxoxoxoxoxox☺", "☺", 22},
   203  	// invalid UTF-8 byte sequence (must be longer than bytealg.MaxBruteForce to
   204  	// test that we don't use IndexRune)
   205  	{"xx0123456789012345678901234567890123456789012345678901234567890120123456789012345678901234567890123456xxx\xed\x9f\xc0", "\xed\x9f\xc0", 105},
   206  }
   207  
   208  var lastIndexTests = []BinOpTest{
   209  	{"", "", 0},
   210  	{"", "a", -1},
   211  	{"", "foo", -1},
   212  	{"fo", "foo", -1},
   213  	{"foo", "foo", 0},
   214  	{"foo", "f", 0},
   215  	{"oofofoofooo", "f", 7},
   216  	{"oofofoofooo", "foo", 7},
   217  	{"barfoobarfoo", "foo", 9},
   218  	{"foo", "", 3},
   219  	{"foo", "o", 2},
   220  	{"abcABCabc", "A", 3},
   221  	{"abcABCabc", "a", 6},
   222  }
   223  
   224  var indexAnyTests = []BinOpTest{
   225  	{"", "", -1},
   226  	{"", "a", -1},
   227  	{"", "abc", -1},
   228  	{"a", "", -1},
   229  	{"a", "a", 0},
   230  	{"\x80", "\xffb", 0},
   231  	{"aaa", "a", 0},
   232  	{"abc", "xyz", -1},
   233  	{"abc", "xcz", 2},
   234  	{"ab☺c", "x☺yz", 2},
   235  	{"a☺b☻c☹d", "cx", len("a☺b☻")},
   236  	{"a☺b☻c☹d", "uvw☻xyz", len("a☺b")},
   237  	{"aRegExp*", ".(|)*+?^$[]", 7},
   238  	{dots + dots + dots, " ", -1},
   239  	{"012abcba210", "\xffb", 4},
   240  	{"012\x80bcb\x80210", "\xffb", 3},
   241  	{"0123456\xcf\x80abc", "\xcfb\x80", 10},
   242  }
   243  
   244  var lastIndexAnyTests = []BinOpTest{
   245  	{"", "", -1},
   246  	{"", "a", -1},
   247  	{"", "abc", -1},
   248  	{"a", "", -1},
   249  	{"a", "a", 0},
   250  	{"\x80", "\xffb", 0},
   251  	{"aaa", "a", 2},
   252  	{"abc", "xyz", -1},
   253  	{"abc", "ab", 1},
   254  	{"ab☺c", "x☺yz", 2},
   255  	{"a☺b☻c☹d", "cx", len("a☺b☻")},
   256  	{"a☺b☻c☹d", "uvw☻xyz", len("a☺b")},
   257  	{"a.RegExp*", ".(|)*+?^$[]", 8},
   258  	{dots + dots + dots, " ", -1},
   259  	{"012abcba210", "\xffb", 6},
   260  	{"012\x80bcb\x80210", "\xffb", 7},
   261  	{"0123456\xcf\x80abc", "\xcfb\x80", 10},
   262  }
   263  
   264  // Execute f on each test case.  funcName should be the name of f; it's used
   265  // in failure reports.
   266  func runIndexTests(t *testing.T, f func(s, sep []byte) int, funcName string, testCases []BinOpTest) {
   267  	for _, test := range testCases {
   268  		a := []byte(test.a)
   269  		b := []byte(test.b)
   270  		actual := f(a, b)
   271  		if actual != test.i {
   272  			t.Errorf("%s(%q,%q) = %v; want %v", funcName, a, b, actual, test.i)
   273  		}
   274  	}
   275  	var allocTests = []struct {
   276  		a []byte
   277  		b []byte
   278  		i int
   279  	}{
   280  		// case for function Index.
   281  		{[]byte("000000000000000000000000000000000000000000000000000000000000000000000001"), []byte("0000000000000000000000000000000000000000000000000000000000000000001"), 5},
   282  		// case for function LastIndex.
   283  		{[]byte("000000000000000000000000000000000000000000000000000000000000000010000"), []byte("00000000000000000000000000000000000000000000000000000000000001"), 3},
   284  	}
   285  	allocs := testing.AllocsPerRun(100, func() {
   286  		if i := Index(allocTests[1].a, allocTests[1].b); i != allocTests[1].i {
   287  			t.Errorf("Index([]byte(%q), []byte(%q)) = %v; want %v", allocTests[1].a, allocTests[1].b, i, allocTests[1].i)
   288  		}
   289  		if i := LastIndex(allocTests[0].a, allocTests[0].b); i != allocTests[0].i {
   290  			t.Errorf("LastIndex([]byte(%q), []byte(%q)) = %v; want %v", allocTests[0].a, allocTests[0].b, i, allocTests[0].i)
   291  		}
   292  	})
   293  	if allocs != 0 {
   294  		t.Errorf("expected no allocations, got %f", allocs)
   295  	}
   296  }
   297  
   298  func runIndexAnyTests(t *testing.T, f func(s []byte, chars string) int, funcName string, testCases []BinOpTest) {
   299  	for _, test := range testCases {
   300  		a := []byte(test.a)
   301  		actual := f(a, test.b)
   302  		if actual != test.i {
   303  			t.Errorf("%s(%q,%q) = %v; want %v", funcName, a, test.b, actual, test.i)
   304  		}
   305  	}
   306  }
   307  
   308  func TestIndex(t *testing.T)     { runIndexTests(t, Index, "Index", indexTests) }
   309  func TestLastIndex(t *testing.T) { runIndexTests(t, LastIndex, "LastIndex", lastIndexTests) }
   310  func TestIndexAny(t *testing.T)  { runIndexAnyTests(t, IndexAny, "IndexAny", indexAnyTests) }
   311  func TestLastIndexAny(t *testing.T) {
   312  	runIndexAnyTests(t, LastIndexAny, "LastIndexAny", lastIndexAnyTests)
   313  }
   314  
   315  func TestIndexByte(t *testing.T) {
   316  	for _, tt := range indexTests {
   317  		if len(tt.b) != 1 {
   318  			continue
   319  		}
   320  		a := []byte(tt.a)
   321  		b := tt.b[0]
   322  		pos := IndexByte(a, b)
   323  		if pos != tt.i {
   324  			t.Errorf(`IndexByte(%q, '%c') = %v`, tt.a, b, pos)
   325  		}
   326  		posp := IndexBytePortable(a, b)
   327  		if posp != tt.i {
   328  			t.Errorf(`indexBytePortable(%q, '%c') = %v`, tt.a, b, posp)
   329  		}
   330  	}
   331  }
   332  
   333  func TestLastIndexByte(t *testing.T) {
   334  	testCases := []BinOpTest{
   335  		{"", "q", -1},
   336  		{"abcdef", "q", -1},
   337  		{"abcdefabcdef", "a", len("abcdef")},      // something in the middle
   338  		{"abcdefabcdef", "f", len("abcdefabcde")}, // last byte
   339  		{"zabcdefabcdef", "z", 0},                 // first byte
   340  		{"a☺b☻c☹d", "b", len("a☺")},               // non-ascii
   341  	}
   342  	for _, test := range testCases {
   343  		actual := LastIndexByte([]byte(test.a), test.b[0])
   344  		if actual != test.i {
   345  			t.Errorf("LastIndexByte(%q,%c) = %v; want %v", test.a, test.b[0], actual, test.i)
   346  		}
   347  	}
   348  }
   349  
   350  // test a larger buffer with different sizes and alignments
   351  func TestIndexByteBig(t *testing.T) {
   352  	var n = 1024
   353  	if testing.Short() {
   354  		n = 128
   355  	}
   356  	b := make([]byte, n)
   357  	for i := 0; i < n; i++ {
   358  		// different start alignments
   359  		b1 := b[i:]
   360  		for j := 0; j < len(b1); j++ {
   361  			b1[j] = 'x'
   362  			pos := IndexByte(b1, 'x')
   363  			if pos != j {
   364  				t.Errorf("IndexByte(%q, 'x') = %v", b1, pos)
   365  			}
   366  			b1[j] = 0
   367  			pos = IndexByte(b1, 'x')
   368  			if pos != -1 {
   369  				t.Errorf("IndexByte(%q, 'x') = %v", b1, pos)
   370  			}
   371  		}
   372  		// different end alignments
   373  		b1 = b[:i]
   374  		for j := 0; j < len(b1); j++ {
   375  			b1[j] = 'x'
   376  			pos := IndexByte(b1, 'x')
   377  			if pos != j {
   378  				t.Errorf("IndexByte(%q, 'x') = %v", b1, pos)
   379  			}
   380  			b1[j] = 0
   381  			pos = IndexByte(b1, 'x')
   382  			if pos != -1 {
   383  				t.Errorf("IndexByte(%q, 'x') = %v", b1, pos)
   384  			}
   385  		}
   386  		// different start and end alignments
   387  		b1 = b[i/2 : n-(i+1)/2]
   388  		for j := 0; j < len(b1); j++ {
   389  			b1[j] = 'x'
   390  			pos := IndexByte(b1, 'x')
   391  			if pos != j {
   392  				t.Errorf("IndexByte(%q, 'x') = %v", b1, pos)
   393  			}
   394  			b1[j] = 0
   395  			pos = IndexByte(b1, 'x')
   396  			if pos != -1 {
   397  				t.Errorf("IndexByte(%q, 'x') = %v", b1, pos)
   398  			}
   399  		}
   400  	}
   401  }
   402  
   403  // test a small index across all page offsets
   404  func TestIndexByteSmall(t *testing.T) {
   405  	b := make([]byte, 5015) // bigger than a page
   406  	// Make sure we find the correct byte even when straddling a page.
   407  	for i := 0; i <= len(b)-15; i++ {
   408  		for j := 0; j < 15; j++ {
   409  			b[i+j] = byte(100 + j)
   410  		}
   411  		for j := 0; j < 15; j++ {
   412  			p := IndexByte(b[i:i+15], byte(100+j))
   413  			if p != j {
   414  				t.Errorf("IndexByte(%q, %d) = %d", b[i:i+15], 100+j, p)
   415  			}
   416  		}
   417  		for j := 0; j < 15; j++ {
   418  			b[i+j] = 0
   419  		}
   420  	}
   421  	// Make sure matches outside the slice never trigger.
   422  	for i := 0; i <= len(b)-15; i++ {
   423  		for j := 0; j < 15; j++ {
   424  			b[i+j] = 1
   425  		}
   426  		for j := 0; j < 15; j++ {
   427  			p := IndexByte(b[i:i+15], byte(0))
   428  			if p != -1 {
   429  				t.Errorf("IndexByte(%q, %d) = %d", b[i:i+15], 0, p)
   430  			}
   431  		}
   432  		for j := 0; j < 15; j++ {
   433  			b[i+j] = 0
   434  		}
   435  	}
   436  }
   437  
   438  func TestIndexRune(t *testing.T) {
   439  	tests := []struct {
   440  		in   string
   441  		rune rune
   442  		want int
   443  	}{
   444  		{"", 'a', -1},
   445  		{"", '☺', -1},
   446  		{"foo", '☹', -1},
   447  		{"foo", 'o', 1},
   448  		{"foo☺bar", '☺', 3},
   449  		{"foo☺☻☹bar", '☹', 9},
   450  		{"a A x", 'A', 2},
   451  		{"some_text=some_value", '=', 9},
   452  		{"☺a", 'a', 3},
   453  		{"a☻☺b", '☺', 4},
   454  		{"𠀳𠀗𠀾𠁄𠀧𠁆𠁂𠀫𠀖𠀪𠀲𠀴𠁀𠀨𠀿", '𠀿', 56},
   455  
   456  		// 2 bytes
   457  		{"ӆ", 'ӆ', 0},
   458  		{"a", 'ӆ', -1},
   459  		{"  ӆ", 'ӆ', 2},
   460  		{"  a", 'ӆ', -1},
   461  		{strings.Repeat("ц", 64) + "ӆ", 'ӆ', 128}, // test cutover
   462  		{strings.Repeat("ц", 64), 'ӆ', -1},
   463  
   464  		// 3 bytes
   465  		{"Ꚁ", 'Ꚁ', 0},
   466  		{"a", 'Ꚁ', -1},
   467  		{"  Ꚁ", 'Ꚁ', 2},
   468  		{"  a", 'Ꚁ', -1},
   469  		{strings.Repeat("Ꙁ", 64) + "Ꚁ", 'Ꚁ', 192}, // test cutover
   470  		{strings.Repeat("Ꙁ", 64) + "Ꚁ", '䚀', -1},  // 'Ꚁ' and '䚀' share the same last two bytes
   471  
   472  		// 4 bytes
   473  		{"𡌀", '𡌀', 0},
   474  		{"a", '𡌀', -1},
   475  		{"  𡌀", '𡌀', 2},
   476  		{"  a", '𡌀', -1},
   477  		{strings.Repeat("𡋀", 64) + "𡌀", '𡌀', 256}, // test cutover
   478  		{strings.Repeat("𡋀", 64) + "𡌀", '𣌀', -1},  // '𡌀' and '𣌀' share the same last two bytes
   479  
   480  		// RuneError should match any invalid UTF-8 byte sequence.
   481  		{"�", '�', 0},
   482  		{"\xff", '�', 0},
   483  		{"☻x�", '�', len("☻x")},
   484  		{"☻x\xe2\x98", '�', len("☻x")},
   485  		{"☻x\xe2\x98�", '�', len("☻x")},
   486  		{"☻x\xe2\x98x", '�', len("☻x")},
   487  
   488  		// Invalid rune values should never match.
   489  		{"a☺b☻c☹d\xe2\x98�\xff�\xed\xa0\x80", -1, -1},
   490  		{"a☺b☻c☹d\xe2\x98�\xff�\xed\xa0\x80", 0xD800, -1}, // Surrogate pair
   491  		{"a☺b☻c☹d\xe2\x98�\xff�\xed\xa0\x80", utf8.MaxRune + 1, -1},
   492  
   493  		// Test the cutover to bytealg.Index when it is triggered in
   494  		// the middle of rune that contains consecutive runs of equal bytes.
   495  		{"aaaaaKKKK\U000bc104", '\U000bc104', 17}, // cutover: (n + 16) / 8
   496  		{"aaaaaKKKK鄄", '鄄', 17},
   497  		{"aaKKKKKa\U000bc104", '\U000bc104', 18}, // cutover: 4 + n>>4
   498  		{"aaKKKKKa鄄", '鄄', 18},
   499  	}
   500  	for _, tt := range tests {
   501  		if got := IndexRune([]byte(tt.in), tt.rune); got != tt.want {
   502  			t.Errorf("IndexRune(%q, %d) = %v; want %v", tt.in, tt.rune, got, tt.want)
   503  		}
   504  	}
   505  
   506  	haystack := []byte("test世界")
   507  	allocs := testing.AllocsPerRun(1000, func() {
   508  		if i := IndexRune(haystack, 's'); i != 2 {
   509  			t.Fatalf("'s' at %d; want 2", i)
   510  		}
   511  		if i := IndexRune(haystack, '世'); i != 4 {
   512  			t.Fatalf("'世' at %d; want 4", i)
   513  		}
   514  	})
   515  	if allocs != 0 {
   516  		t.Errorf("expected no allocations, got %f", allocs)
   517  	}
   518  }
   519  
   520  // test count of a single byte across page offsets
   521  func TestCountByte(t *testing.T) {
   522  	b := make([]byte, 5015) // bigger than a page
   523  	windows := []int{1, 2, 3, 4, 15, 16, 17, 31, 32, 33, 63, 64, 65, 128}
   524  	testCountWindow := func(i, window int) {
   525  		for j := 0; j < window; j++ {
   526  			b[i+j] = byte(100)
   527  			p := Count(b[i:i+window], []byte{100})
   528  			if p != j+1 {
   529  				t.Errorf("TestCountByte.Count(%q, 100) = %d", b[i:i+window], p)
   530  			}
   531  		}
   532  	}
   533  
   534  	maxWnd := windows[len(windows)-1]
   535  
   536  	for i := 0; i <= 2*maxWnd; i++ {
   537  		for _, window := range windows {
   538  			if window > len(b[i:]) {
   539  				window = len(b[i:])
   540  			}
   541  			testCountWindow(i, window)
   542  			for j := 0; j < window; j++ {
   543  				b[i+j] = byte(0)
   544  			}
   545  		}
   546  	}
   547  	for i := 4096 - (maxWnd + 1); i < len(b); i++ {
   548  		for _, window := range windows {
   549  			if window > len(b[i:]) {
   550  				window = len(b[i:])
   551  			}
   552  			testCountWindow(i, window)
   553  			for j := 0; j < window; j++ {
   554  				b[i+j] = byte(0)
   555  			}
   556  		}
   557  	}
   558  }
   559  
   560  // Make sure we don't count bytes outside our window
   561  func TestCountByteNoMatch(t *testing.T) {
   562  	b := make([]byte, 5015)
   563  	windows := []int{1, 2, 3, 4, 15, 16, 17, 31, 32, 33, 63, 64, 65, 128}
   564  	for i := 0; i <= len(b); i++ {
   565  		for _, window := range windows {
   566  			if window > len(b[i:]) {
   567  				window = len(b[i:])
   568  			}
   569  			// Fill the window with non-match
   570  			for j := 0; j < window; j++ {
   571  				b[i+j] = byte(100)
   572  			}
   573  			// Try to find something that doesn't exist
   574  			p := Count(b[i:i+window], []byte{0})
   575  			if p != 0 {
   576  				t.Errorf("TestCountByteNoMatch(%q, 0) = %d", b[i:i+window], p)
   577  			}
   578  			for j := 0; j < window; j++ {
   579  				b[i+j] = byte(0)
   580  			}
   581  		}
   582  	}
   583  }
   584  
   585  var bmbuf []byte
   586  
   587  func valName(x int) string {
   588  	if s := x >> 20; s<<20 == x {
   589  		return fmt.Sprintf("%dM", s)
   590  	}
   591  	if s := x >> 10; s<<10 == x {
   592  		return fmt.Sprintf("%dK", s)
   593  	}
   594  	return fmt.Sprint(x)
   595  }
   596  
   597  func benchBytes(b *testing.B, sizes []int, f func(b *testing.B, n int)) {
   598  	for _, n := range sizes {
   599  		if isRaceBuilder && n > 4<<10 {
   600  			continue
   601  		}
   602  		b.Run(valName(n), func(b *testing.B) {
   603  			if len(bmbuf) < n {
   604  				bmbuf = make([]byte, n)
   605  			}
   606  			b.SetBytes(int64(n))
   607  			f(b, n)
   608  		})
   609  	}
   610  }
   611  
   612  var indexSizes = []int{10, 32, 4 << 10, 4 << 20, 64 << 20}
   613  
   614  var isRaceBuilder = strings.HasSuffix(testenv.Builder(), "-race")
   615  
   616  func BenchmarkIndexByte(b *testing.B) {
   617  	benchBytes(b, indexSizes, bmIndexByte(IndexByte))
   618  }
   619  
   620  func BenchmarkIndexBytePortable(b *testing.B) {
   621  	benchBytes(b, indexSizes, bmIndexByte(IndexBytePortable))
   622  }
   623  
   624  func bmIndexByte(index func([]byte, byte) int) func(b *testing.B, n int) {
   625  	return func(b *testing.B, n int) {
   626  		buf := bmbuf[0:n]
   627  		buf[n-1] = 'x'
   628  		for i := 0; i < b.N; i++ {
   629  			j := index(buf, 'x')
   630  			if j != n-1 {
   631  				b.Fatal("bad index", j)
   632  			}
   633  		}
   634  		buf[n-1] = '\x00'
   635  	}
   636  }
   637  
   638  func BenchmarkIndexRune(b *testing.B) {
   639  	benchBytes(b, indexSizes, bmIndexRune(IndexRune))
   640  }
   641  
   642  func BenchmarkIndexRuneASCII(b *testing.B) {
   643  	benchBytes(b, indexSizes, bmIndexRuneASCII(IndexRune))
   644  }
   645  
   646  func BenchmarkIndexRuneUnicode(b *testing.B) {
   647  	b.Run("Latin", func(b *testing.B) {
   648  		// Latin is mostly 1, 2, 3 byte runes.
   649  		benchBytes(b, indexSizes, bmIndexRuneUnicode(unicode.Latin, 'é'))
   650  	})
   651  	b.Run("Cyrillic", func(b *testing.B) {
   652  		// Cyrillic is mostly 2 and 3 byte runes.
   653  		benchBytes(b, indexSizes, bmIndexRuneUnicode(unicode.Cyrillic, 'Ꙁ'))
   654  	})
   655  	b.Run("Han", func(b *testing.B) {
   656  		// Han consists only of 3 and 4 byte runes.
   657  		benchBytes(b, indexSizes, bmIndexRuneUnicode(unicode.Han, '𠀿'))
   658  	})
   659  }
   660  
   661  func bmIndexRuneASCII(index func([]byte, rune) int) func(b *testing.B, n int) {
   662  	return func(b *testing.B, n int) {
   663  		buf := bmbuf[0:n]
   664  		buf[n-1] = 'x'
   665  		for i := 0; i < b.N; i++ {
   666  			j := index(buf, 'x')
   667  			if j != n-1 {
   668  				b.Fatal("bad index", j)
   669  			}
   670  		}
   671  		buf[n-1] = '\x00'
   672  	}
   673  }
   674  
   675  func bmIndexRune(index func([]byte, rune) int) func(b *testing.B, n int) {
   676  	return func(b *testing.B, n int) {
   677  		buf := bmbuf[0:n]
   678  		utf8.EncodeRune(buf[n-3:], '世')
   679  		for i := 0; i < b.N; i++ {
   680  			j := index(buf, '世')
   681  			if j != n-3 {
   682  				b.Fatal("bad index", j)
   683  			}
   684  		}
   685  		buf[n-3] = '\x00'
   686  		buf[n-2] = '\x00'
   687  		buf[n-1] = '\x00'
   688  	}
   689  }
   690  
   691  func bmIndexRuneUnicode(rt *unicode.RangeTable, needle rune) func(b *testing.B, n int) {
   692  	var rs []rune
   693  	for _, r16 := range rt.R16 {
   694  		for r := rune(r16.Lo); r <= rune(r16.Hi); r += rune(r16.Stride) {
   695  			if r != needle {
   696  				rs = append(rs, rune(r))
   697  			}
   698  		}
   699  	}
   700  	for _, r32 := range rt.R32 {
   701  		for r := rune(r32.Lo); r <= rune(r32.Hi); r += rune(r32.Stride) {
   702  			if r != needle {
   703  				rs = append(rs, rune(r))
   704  			}
   705  		}
   706  	}
   707  	// Shuffle the runes so that they are not in descending order.
   708  	// The sort is deterministic since this is used for benchmarks,
   709  	// which need to be repeatable.
   710  	rr := rand.New(rand.NewSource(1))
   711  	rr.Shuffle(len(rs), func(i, j int) {
   712  		rs[i], rs[j] = rs[j], rs[i]
   713  	})
   714  	uchars := string(rs)
   715  
   716  	return func(b *testing.B, n int) {
   717  		buf := bmbuf[0:n]
   718  		o := copy(buf, uchars)
   719  		for o < len(buf) {
   720  			o += copy(buf[o:], uchars)
   721  		}
   722  
   723  		// Make space for the needle rune at the end of buf.
   724  		m := utf8.RuneLen(needle)
   725  		for o := m; o > 0; {
   726  			_, sz := utf8.DecodeLastRune(buf)
   727  			copy(buf[len(buf)-sz:], "\x00\x00\x00\x00")
   728  			buf = buf[:len(buf)-sz]
   729  			o -= sz
   730  		}
   731  		buf = utf8.AppendRune(buf[:n-m], needle)
   732  
   733  		n -= m // adjust for rune len
   734  		for i := 0; i < b.N; i++ {
   735  			j := IndexRune(buf, needle)
   736  			if j != n {
   737  				b.Fatal("bad index", j)
   738  			}
   739  		}
   740  		for i := range buf {
   741  			buf[i] = '\x00'
   742  		}
   743  	}
   744  }
   745  
   746  func BenchmarkEqual(b *testing.B) {
   747  	b.Run("0", func(b *testing.B) {
   748  		var buf [4]byte
   749  		buf1 := buf[0:0]
   750  		buf2 := buf[1:1]
   751  		for i := 0; i < b.N; i++ {
   752  			eq := Equal(buf1, buf2)
   753  			if !eq {
   754  				b.Fatal("bad equal")
   755  			}
   756  		}
   757  	})
   758  
   759  	sizes := []int{1, 6, 9, 15, 16, 20, 32, 4 << 10, 4 << 20, 64 << 20}
   760  
   761  	b.Run("same", func(b *testing.B) {
   762  		benchBytes(b, sizes, bmEqual(func(a, b []byte) bool { return Equal(a, a) }))
   763  	})
   764  
   765  	benchBytes(b, sizes, bmEqual(Equal))
   766  }
   767  
   768  func bmEqual(equal func([]byte, []byte) bool) func(b *testing.B, n int) {
   769  	return func(b *testing.B, n int) {
   770  		if len(bmbuf) < 2*n {
   771  			bmbuf = make([]byte, 2*n)
   772  		}
   773  		buf1 := bmbuf[0:n]
   774  		buf2 := bmbuf[n : 2*n]
   775  		buf1[n-1] = 'x'
   776  		buf2[n-1] = 'x'
   777  		for i := 0; i < b.N; i++ {
   778  			eq := equal(buf1, buf2)
   779  			if !eq {
   780  				b.Fatal("bad equal")
   781  			}
   782  		}
   783  		buf1[n-1] = '\x00'
   784  		buf2[n-1] = '\x00'
   785  	}
   786  }
   787  
   788  func BenchmarkEqualBothUnaligned(b *testing.B) {
   789  	sizes := []int{64, 4 << 10}
   790  	if !isRaceBuilder {
   791  		sizes = append(sizes, []int{4 << 20, 64 << 20}...)
   792  	}
   793  	maxSize := 2 * (sizes[len(sizes)-1] + 8)
   794  	if len(bmbuf) < maxSize {
   795  		bmbuf = make([]byte, maxSize)
   796  	}
   797  
   798  	for _, n := range sizes {
   799  		for _, off := range []int{0, 1, 4, 7} {
   800  			buf1 := bmbuf[off : off+n]
   801  			buf2Start := (len(bmbuf) / 2) + off
   802  			buf2 := bmbuf[buf2Start : buf2Start+n]
   803  			buf1[n-1] = 'x'
   804  			buf2[n-1] = 'x'
   805  			b.Run(fmt.Sprint(n, off), func(b *testing.B) {
   806  				b.SetBytes(int64(n))
   807  				for i := 0; i < b.N; i++ {
   808  					eq := Equal(buf1, buf2)
   809  					if !eq {
   810  						b.Fatal("bad equal")
   811  					}
   812  				}
   813  			})
   814  			buf1[n-1] = '\x00'
   815  			buf2[n-1] = '\x00'
   816  		}
   817  	}
   818  }
   819  
   820  func BenchmarkIndex(b *testing.B) {
   821  	benchBytes(b, indexSizes, func(b *testing.B, n int) {
   822  		buf := bmbuf[0:n]
   823  		buf[n-1] = 'x'
   824  		for i := 0; i < b.N; i++ {
   825  			j := Index(buf, buf[n-7:])
   826  			if j != n-7 {
   827  				b.Fatal("bad index", j)
   828  			}
   829  		}
   830  		buf[n-1] = '\x00'
   831  	})
   832  }
   833  
   834  func BenchmarkIndexEasy(b *testing.B) {
   835  	benchBytes(b, indexSizes, func(b *testing.B, n int) {
   836  		buf := bmbuf[0:n]
   837  		buf[n-1] = 'x'
   838  		buf[n-7] = 'x'
   839  		for i := 0; i < b.N; i++ {
   840  			j := Index(buf, buf[n-7:])
   841  			if j != n-7 {
   842  				b.Fatal("bad index", j)
   843  			}
   844  		}
   845  		buf[n-1] = '\x00'
   846  		buf[n-7] = '\x00'
   847  	})
   848  }
   849  
   850  func BenchmarkCount(b *testing.B) {
   851  	benchBytes(b, indexSizes, func(b *testing.B, n int) {
   852  		buf := bmbuf[0:n]
   853  		buf[n-1] = 'x'
   854  		for i := 0; i < b.N; i++ {
   855  			j := Count(buf, buf[n-7:])
   856  			if j != 1 {
   857  				b.Fatal("bad count", j)
   858  			}
   859  		}
   860  		buf[n-1] = '\x00'
   861  	})
   862  }
   863  
   864  func BenchmarkCountEasy(b *testing.B) {
   865  	benchBytes(b, indexSizes, func(b *testing.B, n int) {
   866  		buf := bmbuf[0:n]
   867  		buf[n-1] = 'x'
   868  		buf[n-7] = 'x'
   869  		for i := 0; i < b.N; i++ {
   870  			j := Count(buf, buf[n-7:])
   871  			if j != 1 {
   872  				b.Fatal("bad count", j)
   873  			}
   874  		}
   875  		buf[n-1] = '\x00'
   876  		buf[n-7] = '\x00'
   877  	})
   878  }
   879  
   880  func BenchmarkCountSingle(b *testing.B) {
   881  	benchBytes(b, indexSizes, func(b *testing.B, n int) {
   882  		buf := bmbuf[0:n]
   883  		step := 8
   884  		for i := 0; i < len(buf); i += step {
   885  			buf[i] = 1
   886  		}
   887  		expect := (len(buf) + (step - 1)) / step
   888  		for i := 0; i < b.N; i++ {
   889  			j := Count(buf, []byte{1})
   890  			if j != expect {
   891  				b.Fatal("bad count", j, expect)
   892  			}
   893  		}
   894  		for i := 0; i < len(buf); i++ {
   895  			buf[i] = 0
   896  		}
   897  	})
   898  }
   899  
   900  type SplitTest struct {
   901  	s   string
   902  	sep string
   903  	n   int
   904  	a   []string
   905  }
   906  
   907  var splittests = []SplitTest{
   908  	{"", "", -1, []string{}},
   909  	{abcd, "a", 0, nil},
   910  	{abcd, "", 2, []string{"a", "bcd"}},
   911  	{abcd, "a", -1, []string{"", "bcd"}},
   912  	{abcd, "z", -1, []string{"abcd"}},
   913  	{abcd, "", -1, []string{"a", "b", "c", "d"}},
   914  	{commas, ",", -1, []string{"1", "2", "3", "4"}},
   915  	{dots, "...", -1, []string{"1", ".2", ".3", ".4"}},
   916  	{faces, "☹", -1, []string{"☺☻", ""}},
   917  	{faces, "~", -1, []string{faces}},
   918  	{faces, "", -1, []string{"☺", "☻", "☹"}},
   919  	{"1 2 3 4", " ", 3, []string{"1", "2", "3 4"}},
   920  	{"1 2", " ", 3, []string{"1", "2"}},
   921  	{"123", "", 2, []string{"1", "23"}},
   922  	{"123", "", 17, []string{"1", "2", "3"}},
   923  	{"bT", "T", math.MaxInt / 4, []string{"b", ""}},
   924  	{"\xff-\xff", "", -1, []string{"\xff", "-", "\xff"}},
   925  	{"\xff-\xff", "-", -1, []string{"\xff", "\xff"}},
   926  }
   927  
   928  func TestSplit(t *testing.T) {
   929  	for _, tt := range splittests {
   930  		a := SplitN([]byte(tt.s), []byte(tt.sep), tt.n)
   931  
   932  		// Appending to the results should not change future results.
   933  		var x []byte
   934  		for _, v := range a {
   935  			x = append(v, 'z')
   936  		}
   937  
   938  		result := sliceOfString(a)
   939  		if !slices.Equal(result, tt.a) {
   940  			t.Errorf(`Split(%q, %q, %d) = %v; want %v`, tt.s, tt.sep, tt.n, result, tt.a)
   941  			continue
   942  		}
   943  
   944  		if tt.n < 0 {
   945  			b := sliceOfString(slices.Collect(SplitSeq([]byte(tt.s), []byte(tt.sep))))
   946  			if !slices.Equal(b, tt.a) {
   947  				t.Errorf(`collect(SplitSeq(%q, %q)) = %v; want %v`, tt.s, tt.sep, b, tt.a)
   948  			}
   949  		}
   950  
   951  		if tt.n == 0 || len(a) == 0 {
   952  			continue
   953  		}
   954  
   955  		if want := tt.a[len(tt.a)-1] + "z"; string(x) != want {
   956  			t.Errorf("last appended result was %s; want %s", x, want)
   957  		}
   958  
   959  		s := Join(a, []byte(tt.sep))
   960  		if string(s) != tt.s {
   961  			t.Errorf(`Join(Split(%q, %q, %d), %q) = %q`, tt.s, tt.sep, tt.n, tt.sep, s)
   962  		}
   963  		if tt.n < 0 {
   964  			b := sliceOfString(Split([]byte(tt.s), []byte(tt.sep)))
   965  			if !slices.Equal(result, b) {
   966  				t.Errorf("Split disagrees withSplitN(%q, %q, %d) = %v; want %v", tt.s, tt.sep, tt.n, b, a)
   967  			}
   968  		}
   969  		if len(a) > 0 {
   970  			in, out := a[0], s
   971  			if cap(in) == cap(out) && &in[:1][0] == &out[:1][0] {
   972  				t.Errorf("Join(%#v, %q) didn't copy", a, tt.sep)
   973  			}
   974  		}
   975  	}
   976  }
   977  
   978  var splitaftertests = []SplitTest{
   979  	{abcd, "a", -1, []string{"a", "bcd"}},
   980  	{abcd, "z", -1, []string{"abcd"}},
   981  	{abcd, "", -1, []string{"a", "b", "c", "d"}},
   982  	{commas, ",", -1, []string{"1,", "2,", "3,", "4"}},
   983  	{dots, "...", -1, []string{"1...", ".2...", ".3...", ".4"}},
   984  	{faces, "☹", -1, []string{"☺☻☹", ""}},
   985  	{faces, "~", -1, []string{faces}},
   986  	{faces, "", -1, []string{"☺", "☻", "☹"}},
   987  	{"1 2 3 4", " ", 3, []string{"1 ", "2 ", "3 4"}},
   988  	{"1 2 3", " ", 3, []string{"1 ", "2 ", "3"}},
   989  	{"1 2", " ", 3, []string{"1 ", "2"}},
   990  	{"123", "", 2, []string{"1", "23"}},
   991  	{"123", "", 17, []string{"1", "2", "3"}},
   992  }
   993  
   994  func TestSplitAfter(t *testing.T) {
   995  	for _, tt := range splitaftertests {
   996  		a := SplitAfterN([]byte(tt.s), []byte(tt.sep), tt.n)
   997  
   998  		// Appending to the results should not change future results.
   999  		var x []byte
  1000  		for _, v := range a {
  1001  			x = append(v, 'z')
  1002  		}
  1003  
  1004  		result := sliceOfString(a)
  1005  		if !slices.Equal(result, tt.a) {
  1006  			t.Errorf(`Split(%q, %q, %d) = %v; want %v`, tt.s, tt.sep, tt.n, result, tt.a)
  1007  			continue
  1008  		}
  1009  
  1010  		if tt.n < 0 {
  1011  			b := sliceOfString(slices.Collect(SplitAfterSeq([]byte(tt.s), []byte(tt.sep))))
  1012  			if !slices.Equal(b, tt.a) {
  1013  				t.Errorf(`collect(SplitAfterSeq(%q, %q)) = %v; want %v`, tt.s, tt.sep, b, tt.a)
  1014  			}
  1015  		}
  1016  
  1017  		if want := tt.a[len(tt.a)-1] + "z"; string(x) != want {
  1018  			t.Errorf("last appended result was %s; want %s", x, want)
  1019  		}
  1020  
  1021  		s := Join(a, nil)
  1022  		if string(s) != tt.s {
  1023  			t.Errorf(`Join(Split(%q, %q, %d), %q) = %q`, tt.s, tt.sep, tt.n, tt.sep, s)
  1024  		}
  1025  		if tt.n < 0 {
  1026  			b := sliceOfString(SplitAfter([]byte(tt.s), []byte(tt.sep)))
  1027  			if !slices.Equal(result, b) {
  1028  				t.Errorf("SplitAfter disagrees withSplitAfterN(%q, %q, %d) = %v; want %v", tt.s, tt.sep, tt.n, b, a)
  1029  			}
  1030  		}
  1031  	}
  1032  }
  1033  
  1034  type FieldsTest struct {
  1035  	s string
  1036  	a []string
  1037  }
  1038  
  1039  var fieldstests = []FieldsTest{
  1040  	{"", []string{}},
  1041  	{" ", []string{}},
  1042  	{" \t ", []string{}},
  1043  	{"  abc  ", []string{"abc"}},
  1044  	{"1 2 3 4", []string{"1", "2", "3", "4"}},
  1045  	{"1  2  3  4", []string{"1", "2", "3", "4"}},
  1046  	{"1\t\t2\t\t3\t4", []string{"1", "2", "3", "4"}},
  1047  	{"1\u20002\u20013\u20024", []string{"1", "2", "3", "4"}},
  1048  	{"\u2000\u2001\u2002", []string{}},
  1049  	{"\n™\t™\n", []string{"™", "™"}},
  1050  	{faces, []string{faces}},
  1051  }
  1052  
  1053  func TestFields(t *testing.T) {
  1054  	for _, tt := range fieldstests {
  1055  		b := []byte(tt.s)
  1056  		a := Fields(b)
  1057  
  1058  		// Appending to the results should not change future results.
  1059  		var x []byte
  1060  		for _, v := range a {
  1061  			x = append(v, 'z')
  1062  		}
  1063  
  1064  		result := sliceOfString(a)
  1065  		if !slices.Equal(result, tt.a) {
  1066  			t.Errorf("Fields(%q) = %v; want %v", tt.s, a, tt.a)
  1067  			continue
  1068  		}
  1069  
  1070  		result2 := sliceOfString(collect(t, FieldsSeq([]byte(tt.s))))
  1071  		if !slices.Equal(result2, tt.a) {
  1072  			t.Errorf(`collect(FieldsSeq(%q)) = %v; want %v`, tt.s, result2, tt.a)
  1073  		}
  1074  
  1075  		if string(b) != tt.s {
  1076  			t.Errorf("slice changed to %s; want %s", string(b), tt.s)
  1077  		}
  1078  		if len(tt.a) > 0 {
  1079  			if want := tt.a[len(tt.a)-1] + "z"; string(x) != want {
  1080  				t.Errorf("last appended result was %s; want %s", x, want)
  1081  			}
  1082  		}
  1083  	}
  1084  }
  1085  
  1086  func TestFieldsFunc(t *testing.T) {
  1087  	for _, tt := range fieldstests {
  1088  		a := FieldsFunc([]byte(tt.s), unicode.IsSpace)
  1089  		result := sliceOfString(a)
  1090  		if !slices.Equal(result, tt.a) {
  1091  			t.Errorf("FieldsFunc(%q, unicode.IsSpace) = %v; want %v", tt.s, a, tt.a)
  1092  			continue
  1093  		}
  1094  	}
  1095  	pred := func(c rune) bool { return c == 'X' }
  1096  	var fieldsFuncTests = []FieldsTest{
  1097  		{"", []string{}},
  1098  		{"XX", []string{}},
  1099  		{"XXhiXXX", []string{"hi"}},
  1100  		{"aXXbXXXcX", []string{"a", "b", "c"}},
  1101  	}
  1102  	for _, tt := range fieldsFuncTests {
  1103  		b := []byte(tt.s)
  1104  		a := FieldsFunc(b, pred)
  1105  
  1106  		// Appending to the results should not change future results.
  1107  		var x []byte
  1108  		for _, v := range a {
  1109  			x = append(v, 'z')
  1110  		}
  1111  
  1112  		result := sliceOfString(a)
  1113  		if !slices.Equal(result, tt.a) {
  1114  			t.Errorf("FieldsFunc(%q) = %v, want %v", tt.s, a, tt.a)
  1115  		}
  1116  
  1117  		result2 := sliceOfString(collect(t, FieldsFuncSeq([]byte(tt.s), pred)))
  1118  		if !slices.Equal(result2, tt.a) {
  1119  			t.Errorf(`collect(FieldsFuncSeq(%q)) = %v; want %v`, tt.s, result2, tt.a)
  1120  		}
  1121  
  1122  		if string(b) != tt.s {
  1123  			t.Errorf("slice changed to %s; want %s", b, tt.s)
  1124  		}
  1125  		if len(tt.a) > 0 {
  1126  			if want := tt.a[len(tt.a)-1] + "z"; string(x) != want {
  1127  				t.Errorf("last appended result was %s; want %s", x, want)
  1128  			}
  1129  		}
  1130  	}
  1131  }
  1132  
  1133  // Test case for any function which accepts and returns a byte slice.
  1134  // For ease of creation, we write the input byte slice as a string.
  1135  type StringTest struct {
  1136  	in  string
  1137  	out []byte
  1138  }
  1139  
  1140  var upperTests = []StringTest{
  1141  	{"", []byte("")},
  1142  	{"ONLYUPPER", []byte("ONLYUPPER")},
  1143  	{"abc", []byte("ABC")},
  1144  	{"AbC123", []byte("ABC123")},
  1145  	{"azAZ09_", []byte("AZAZ09_")},
  1146  	{"longStrinGwitHmixofsmaLLandcAps", []byte("LONGSTRINGWITHMIXOFSMALLANDCAPS")},
  1147  	{"long\u0250string\u0250with\u0250nonascii\u2C6Fchars", []byte("LONG\u2C6FSTRING\u2C6FWITH\u2C6FNONASCII\u2C6FCHARS")},
  1148  	{"\u0250\u0250\u0250\u0250\u0250", []byte("\u2C6F\u2C6F\u2C6F\u2C6F\u2C6F")}, // grows one byte per char
  1149  	{"a\u0080\U0010FFFF", []byte("A\u0080\U0010FFFF")},                           // test utf8.RuneSelf and utf8.MaxRune
  1150  }
  1151  
  1152  var lowerTests = []StringTest{
  1153  	{"", []byte("")},
  1154  	{"abc", []byte("abc")},
  1155  	{"AbC123", []byte("abc123")},
  1156  	{"azAZ09_", []byte("azaz09_")},
  1157  	{"longStrinGwitHmixofsmaLLandcAps", []byte("longstringwithmixofsmallandcaps")},
  1158  	{"LONG\u2C6FSTRING\u2C6FWITH\u2C6FNONASCII\u2C6FCHARS", []byte("long\u0250string\u0250with\u0250nonascii\u0250chars")},
  1159  	{"\u2C6D\u2C6D\u2C6D\u2C6D\u2C6D", []byte("\u0251\u0251\u0251\u0251\u0251")}, // shrinks one byte per char
  1160  	{"A\u0080\U0010FFFF", []byte("a\u0080\U0010FFFF")},                           // test utf8.RuneSelf and utf8.MaxRune
  1161  }
  1162  
  1163  const space = "\t\v\r\f\n\u0085\u00a0\u2000\u3000"
  1164  
  1165  var trimSpaceTests = []StringTest{
  1166  	{"", nil},
  1167  	{"  a", []byte("a")},
  1168  	{"b  ", []byte("b")},
  1169  	{"abc", []byte("abc")},
  1170  	{space + "abc" + space, []byte("abc")},
  1171  	{" ", nil},
  1172  	{"\u3000 ", nil},
  1173  	{" \u3000", nil},
  1174  	{" \t\r\n \t\t\r\r\n\n ", nil},
  1175  	{" \t\r\n x\t\t\r\r\n\n ", []byte("x")},
  1176  	{" \u2000\t\r\n x\t\t\r\r\ny\n \u3000", []byte("x\t\t\r\r\ny")},
  1177  	{"1 \t\r\n2", []byte("1 \t\r\n2")},
  1178  	{" x\x80", []byte("x\x80")},
  1179  	{" x\xc0", []byte("x\xc0")},
  1180  	{"x \xc0\xc0 ", []byte("x \xc0\xc0")},
  1181  	{"x \xc0", []byte("x \xc0")},
  1182  	{"x \xc0 ", []byte("x \xc0")},
  1183  	{"x \xc0\xc0 ", []byte("x \xc0\xc0")},
  1184  	{"x ☺\xc0\xc0 ", []byte("x ☺\xc0\xc0")},
  1185  	{"x ☺ ", []byte("x ☺")},
  1186  }
  1187  
  1188  // Execute f on each test case.  funcName should be the name of f; it's used
  1189  // in failure reports.
  1190  func runStringTests(t *testing.T, f func([]byte) []byte, funcName string, testCases []StringTest) {
  1191  	for _, tc := range testCases {
  1192  		actual := f([]byte(tc.in))
  1193  		if actual == nil && tc.out != nil {
  1194  			t.Errorf("%s(%q) = nil; want %q", funcName, tc.in, tc.out)
  1195  		}
  1196  		if actual != nil && tc.out == nil {
  1197  			t.Errorf("%s(%q) = %q; want nil", funcName, tc.in, actual)
  1198  		}
  1199  		if !Equal(actual, tc.out) {
  1200  			t.Errorf("%s(%q) = %q; want %q", funcName, tc.in, actual, tc.out)
  1201  		}
  1202  	}
  1203  }
  1204  
  1205  func tenRunes(r rune) string {
  1206  	runes := make([]rune, 10)
  1207  	for i := range runes {
  1208  		runes[i] = r
  1209  	}
  1210  	return string(runes)
  1211  }
  1212  
  1213  // User-defined self-inverse mapping function
  1214  func rot13(r rune) rune {
  1215  	const step = 13
  1216  	if r >= 'a' && r <= 'z' {
  1217  		return ((r - 'a' + step) % 26) + 'a'
  1218  	}
  1219  	if r >= 'A' && r <= 'Z' {
  1220  		return ((r - 'A' + step) % 26) + 'A'
  1221  	}
  1222  	return r
  1223  }
  1224  
  1225  func TestMap(t *testing.T) {
  1226  	// Run a couple of awful growth/shrinkage tests
  1227  	a := tenRunes('a')
  1228  
  1229  	// 1.  Grow. This triggers two reallocations in Map.
  1230  	maxRune := func(r rune) rune { return unicode.MaxRune }
  1231  	m := Map(maxRune, []byte(a))
  1232  	expect := tenRunes(unicode.MaxRune)
  1233  	if string(m) != expect {
  1234  		t.Errorf("growing: expected %q got %q", expect, m)
  1235  	}
  1236  
  1237  	// 2. Shrink
  1238  	minRune := func(r rune) rune { return 'a' }
  1239  	m = Map(minRune, []byte(tenRunes(unicode.MaxRune)))
  1240  	expect = a
  1241  	if string(m) != expect {
  1242  		t.Errorf("shrinking: expected %q got %q", expect, m)
  1243  	}
  1244  
  1245  	// 3. Rot13
  1246  	m = Map(rot13, []byte("a to zed"))
  1247  	expect = "n gb mrq"
  1248  	if string(m) != expect {
  1249  		t.Errorf("rot13: expected %q got %q", expect, m)
  1250  	}
  1251  
  1252  	// 4. Rot13^2
  1253  	m = Map(rot13, Map(rot13, []byte("a to zed")))
  1254  	expect = "a to zed"
  1255  	if string(m) != expect {
  1256  		t.Errorf("rot13: expected %q got %q", expect, m)
  1257  	}
  1258  
  1259  	// 5. Drop
  1260  	dropNotLatin := func(r rune) rune {
  1261  		if unicode.Is(unicode.Latin, r) {
  1262  			return r
  1263  		}
  1264  		return -1
  1265  	}
  1266  	m = Map(dropNotLatin, []byte("Hello, 세계"))
  1267  	expect = "Hello"
  1268  	if string(m) != expect {
  1269  		t.Errorf("drop: expected %q got %q", expect, m)
  1270  	}
  1271  
  1272  	// 6. Invalid rune
  1273  	invalidRune := func(r rune) rune {
  1274  		return utf8.MaxRune + 1
  1275  	}
  1276  	m = Map(invalidRune, []byte("x"))
  1277  	expect = "\uFFFD"
  1278  	if string(m) != expect {
  1279  		t.Errorf("invalidRune: expected %q got %q", expect, m)
  1280  	}
  1281  }
  1282  
  1283  func TestToUpper(t *testing.T) { runStringTests(t, ToUpper, "ToUpper", upperTests) }
  1284  
  1285  func TestToLower(t *testing.T) { runStringTests(t, ToLower, "ToLower", lowerTests) }
  1286  
  1287  func BenchmarkToUpper(b *testing.B) {
  1288  	for _, tc := range upperTests {
  1289  		tin := []byte(tc.in)
  1290  		b.Run(tc.in, func(b *testing.B) {
  1291  			for i := 0; i < b.N; i++ {
  1292  				actual := ToUpper(tin)
  1293  				if !Equal(actual, tc.out) {
  1294  					b.Errorf("ToUpper(%q) = %q; want %q", tc.in, actual, tc.out)
  1295  				}
  1296  			}
  1297  		})
  1298  	}
  1299  }
  1300  
  1301  func BenchmarkToLower(b *testing.B) {
  1302  	for _, tc := range lowerTests {
  1303  		tin := []byte(tc.in)
  1304  		b.Run(tc.in, func(b *testing.B) {
  1305  			for i := 0; i < b.N; i++ {
  1306  				actual := ToLower(tin)
  1307  				if !Equal(actual, tc.out) {
  1308  					b.Errorf("ToLower(%q) = %q; want %q", tc.in, actual, tc.out)
  1309  				}
  1310  			}
  1311  		})
  1312  	}
  1313  }
  1314  
  1315  var toValidUTF8Tests = []struct {
  1316  	in   string
  1317  	repl string
  1318  	out  string
  1319  }{
  1320  	{"", "\uFFFD", ""},
  1321  	{"abc", "\uFFFD", "abc"},
  1322  	{"\uFDDD", "\uFFFD", "\uFDDD"},
  1323  	{"a\xffb", "\uFFFD", "a\uFFFDb"},
  1324  	{"a\xffb\uFFFD", "X", "aXb\uFFFD"},
  1325  	{"a☺\xffb☺\xC0\xAFc☺\xff", "", "a☺b☺c☺"},
  1326  	{"a☺\xffb☺\xC0\xAFc☺\xff", "日本語", "a☺日本語b☺日本語c☺日本語"},
  1327  	{"\xC0\xAF", "\uFFFD", "\uFFFD"},
  1328  	{"\xE0\x80\xAF", "\uFFFD", "\uFFFD"},
  1329  	{"\xed\xa0\x80", "abc", "abc"},
  1330  	{"\xed\xbf\xbf", "\uFFFD", "\uFFFD"},
  1331  	{"\xF0\x80\x80\xaf", "☺", "☺"},
  1332  	{"\xF8\x80\x80\x80\xAF", "\uFFFD", "\uFFFD"},
  1333  	{"\xFC\x80\x80\x80\x80\xAF", "\uFFFD", "\uFFFD"},
  1334  }
  1335  
  1336  func TestToValidUTF8(t *testing.T) {
  1337  	for _, tc := range toValidUTF8Tests {
  1338  		got := ToValidUTF8([]byte(tc.in), []byte(tc.repl))
  1339  		if !Equal(got, []byte(tc.out)) {
  1340  			t.Errorf("ToValidUTF8(%q, %q) = %q; want %q", tc.in, tc.repl, got, tc.out)
  1341  		}
  1342  	}
  1343  }
  1344  
  1345  func TestTrimSpace(t *testing.T) { runStringTests(t, TrimSpace, "TrimSpace", trimSpaceTests) }
  1346  
  1347  type RepeatTest struct {
  1348  	in, out string
  1349  	count   int
  1350  }
  1351  
  1352  var longString = "a" + string(make([]byte, 1<<16)) + "z"
  1353  
  1354  var RepeatTests = []RepeatTest{
  1355  	{"", "", 0},
  1356  	{"", "", 1},
  1357  	{"", "", 2},
  1358  	{"-", "", 0},
  1359  	{"-", "-", 1},
  1360  	{"-", "----------", 10},
  1361  	{"abc ", "abc abc abc ", 3},
  1362  	// Tests for results over the chunkLimit
  1363  	{string(rune(0)), string(make([]byte, 1<<16)), 1 << 16},
  1364  	{longString, longString + longString, 2},
  1365  }
  1366  
  1367  func TestRepeat(t *testing.T) {
  1368  	for _, tt := range RepeatTests {
  1369  		tin := []byte(tt.in)
  1370  		tout := []byte(tt.out)
  1371  		a := Repeat(tin, tt.count)
  1372  		if !Equal(a, tout) {
  1373  			t.Errorf("Repeat(%q, %d) = %q; want %q", tin, tt.count, a, tout)
  1374  			continue
  1375  		}
  1376  	}
  1377  }
  1378  
  1379  func repeat(b []byte, count int) (err error) {
  1380  	defer func() {
  1381  		if r := recover(); r != nil {
  1382  			switch v := r.(type) {
  1383  			case error:
  1384  				err = v
  1385  			default:
  1386  				err = fmt.Errorf("%s", v)
  1387  			}
  1388  		}
  1389  	}()
  1390  
  1391  	Repeat(b, count)
  1392  
  1393  	return
  1394  }
  1395  
  1396  // See Issue golang.org/issue/16237
  1397  func TestRepeatCatchesOverflow(t *testing.T) {
  1398  	type testCase struct {
  1399  		s      string
  1400  		count  int
  1401  		errStr string
  1402  	}
  1403  
  1404  	runTestCases := func(prefix string, tests []testCase) {
  1405  		for i, tt := range tests {
  1406  			err := repeat([]byte(tt.s), tt.count)
  1407  			if tt.errStr == "" {
  1408  				if err != nil {
  1409  					t.Errorf("#%d panicked %v", i, err)
  1410  				}
  1411  				continue
  1412  			}
  1413  
  1414  			if err == nil || !strings.Contains(err.Error(), tt.errStr) {
  1415  				t.Errorf("%s#%d got %q want %q", prefix, i, err, tt.errStr)
  1416  			}
  1417  		}
  1418  	}
  1419  
  1420  	const maxInt = int(^uint(0) >> 1)
  1421  
  1422  	runTestCases("", []testCase{
  1423  		0: {"--", -2147483647, "negative"},
  1424  		1: {"", maxInt, ""},
  1425  		2: {"-", 10, ""},
  1426  		3: {"gopher", 0, ""},
  1427  		4: {"-", -1, "negative"},
  1428  		5: {"--", -102, "negative"},
  1429  		6: {string(make([]byte, 255)), int((^uint(0))/255 + 1), "overflow"},
  1430  	})
  1431  
  1432  	const is64Bit = 1<<(^uintptr(0)>>63)/2 != 0
  1433  	if !is64Bit {
  1434  		return
  1435  	}
  1436  
  1437  	runTestCases("64-bit", []testCase{
  1438  		0: {"-", maxInt, "out of range"},
  1439  	})
  1440  }
  1441  
  1442  type RunesTest struct {
  1443  	in    string
  1444  	out   []rune
  1445  	lossy bool
  1446  }
  1447  
  1448  var RunesTests = []RunesTest{
  1449  	{"", []rune{}, false},
  1450  	{" ", []rune{32}, false},
  1451  	{"ABC", []rune{65, 66, 67}, false},
  1452  	{"abc", []rune{97, 98, 99}, false},
  1453  	{"\u65e5\u672c\u8a9e", []rune{26085, 26412, 35486}, false},
  1454  	{"ab\x80c", []rune{97, 98, 0xFFFD, 99}, true},
  1455  	{"ab\xc0c", []rune{97, 98, 0xFFFD, 99}, true},
  1456  }
  1457  
  1458  func TestRunes(t *testing.T) {
  1459  	for _, tt := range RunesTests {
  1460  		tin := []byte(tt.in)
  1461  		a := Runes(tin)
  1462  		if !slices.Equal(a, tt.out) {
  1463  			t.Errorf("Runes(%q) = %v; want %v", tin, a, tt.out)
  1464  			continue
  1465  		}
  1466  		if !tt.lossy {
  1467  			// can only test reassembly if we didn't lose information
  1468  			s := string(a)
  1469  			if s != tt.in {
  1470  				t.Errorf("string(Runes(%q)) = %x; want %x", tin, s, tin)
  1471  			}
  1472  		}
  1473  	}
  1474  }
  1475  
  1476  type TrimTest struct {
  1477  	f            string
  1478  	in, arg, out string
  1479  }
  1480  
  1481  var trimTests = []TrimTest{
  1482  	{"Trim", "abba", "a", "bb"},
  1483  	{"Trim", "abba", "ab", ""},
  1484  	{"TrimLeft", "abba", "ab", ""},
  1485  	{"TrimRight", "abba", "ab", ""},
  1486  	{"TrimLeft", "abba", "a", "bba"},
  1487  	{"TrimLeft", "abba", "b", "abba"},
  1488  	{"TrimRight", "abba", "a", "abb"},
  1489  	{"TrimRight", "abba", "b", "abba"},
  1490  	{"Trim", "<tag>", "<>", "tag"},
  1491  	{"Trim", "* listitem", " *", "listitem"},
  1492  	{"Trim", `"quote"`, `"`, "quote"},
  1493  	{"Trim", "\u2C6F\u2C6F\u0250\u0250\u2C6F\u2C6F", "\u2C6F", "\u0250\u0250"},
  1494  	{"Trim", "\x80test\xff", "\xff", "test"},
  1495  	{"Trim", " Ġ ", " ", "Ġ"},
  1496  	{"Trim", " Ġİ0", "0 ", "Ġİ"},
  1497  	//empty string tests
  1498  	{"Trim", "abba", "", "abba"},
  1499  	{"Trim", "", "123", ""},
  1500  	{"Trim", "", "", ""},
  1501  	{"TrimLeft", "abba", "", "abba"},
  1502  	{"TrimLeft", "", "123", ""},
  1503  	{"TrimLeft", "", "", ""},
  1504  	{"TrimRight", "abba", "", "abba"},
  1505  	{"TrimRight", "", "123", ""},
  1506  	{"TrimRight", "", "", ""},
  1507  	{"TrimRight", "☺\xc0", "☺", "☺\xc0"},
  1508  	{"TrimPrefix", "aabb", "a", "abb"},
  1509  	{"TrimPrefix", "aabb", "b", "aabb"},
  1510  	{"TrimSuffix", "aabb", "a", "aabb"},
  1511  	{"TrimSuffix", "aabb", "b", "aab"},
  1512  }
  1513  
  1514  type TrimNilTest struct {
  1515  	f   string
  1516  	in  []byte
  1517  	arg string
  1518  	out []byte
  1519  }
  1520  
  1521  var trimNilTests = []TrimNilTest{
  1522  	{"Trim", nil, "", nil},
  1523  	{"Trim", []byte{}, "", nil},
  1524  	{"Trim", []byte{'a'}, "a", nil},
  1525  	{"Trim", []byte{'a', 'a'}, "a", nil},
  1526  	{"Trim", []byte{'a'}, "ab", nil},
  1527  	{"Trim", []byte{'a', 'b'}, "ab", nil},
  1528  	{"Trim", []byte("☺"), "☺", nil},
  1529  	{"TrimLeft", nil, "", nil},
  1530  	{"TrimLeft", []byte{}, "", nil},
  1531  	{"TrimLeft", []byte{'a'}, "a", nil},
  1532  	{"TrimLeft", []byte{'a', 'a'}, "a", nil},
  1533  	{"TrimLeft", []byte{'a'}, "ab", nil},
  1534  	{"TrimLeft", []byte{'a', 'b'}, "ab", nil},
  1535  	{"TrimLeft", []byte("☺"), "☺", nil},
  1536  	{"TrimRight", nil, "", nil},
  1537  	{"TrimRight", []byte{}, "", []byte{}},
  1538  	{"TrimRight", []byte{'a'}, "a", []byte{}},
  1539  	{"TrimRight", []byte{'a', 'a'}, "a", []byte{}},
  1540  	{"TrimRight", []byte{'a'}, "ab", []byte{}},
  1541  	{"TrimRight", []byte{'a', 'b'}, "ab", []byte{}},
  1542  	{"TrimRight", []byte("☺"), "☺", []byte{}},
  1543  	{"TrimPrefix", nil, "", nil},
  1544  	{"TrimPrefix", []byte{}, "", []byte{}},
  1545  	{"TrimPrefix", []byte{'a'}, "a", []byte{}},
  1546  	{"TrimPrefix", []byte("☺"), "☺", []byte{}},
  1547  	{"TrimSuffix", nil, "", nil},
  1548  	{"TrimSuffix", []byte{}, "", []byte{}},
  1549  	{"TrimSuffix", []byte{'a'}, "a", []byte{}},
  1550  	{"TrimSuffix", []byte("☺"), "☺", []byte{}},
  1551  }
  1552  
  1553  func TestTrim(t *testing.T) {
  1554  	toFn := func(name string) (func([]byte, string) []byte, func([]byte, []byte) []byte) {
  1555  		switch name {
  1556  		case "Trim":
  1557  			return Trim, nil
  1558  		case "TrimLeft":
  1559  			return TrimLeft, nil
  1560  		case "TrimRight":
  1561  			return TrimRight, nil
  1562  		case "TrimPrefix":
  1563  			return nil, TrimPrefix
  1564  		case "TrimSuffix":
  1565  			return nil, TrimSuffix
  1566  		default:
  1567  			t.Errorf("Undefined trim function %s", name)
  1568  			return nil, nil
  1569  		}
  1570  	}
  1571  
  1572  	for _, tc := range trimTests {
  1573  		name := tc.f
  1574  		f, fb := toFn(name)
  1575  		if f == nil && fb == nil {
  1576  			continue
  1577  		}
  1578  		var actual string
  1579  		if f != nil {
  1580  			actual = string(f([]byte(tc.in), tc.arg))
  1581  		} else {
  1582  			actual = string(fb([]byte(tc.in), []byte(tc.arg)))
  1583  		}
  1584  		if actual != tc.out {
  1585  			t.Errorf("%s(%q, %q) = %q; want %q", name, tc.in, tc.arg, actual, tc.out)
  1586  		}
  1587  	}
  1588  
  1589  	for _, tc := range trimNilTests {
  1590  		name := tc.f
  1591  		f, fb := toFn(name)
  1592  		if f == nil && fb == nil {
  1593  			continue
  1594  		}
  1595  		var actual []byte
  1596  		if f != nil {
  1597  			actual = f(tc.in, tc.arg)
  1598  		} else {
  1599  			actual = fb(tc.in, []byte(tc.arg))
  1600  		}
  1601  		report := func(s []byte) string {
  1602  			if s == nil {
  1603  				return "nil"
  1604  			} else {
  1605  				return fmt.Sprintf("%q", s)
  1606  			}
  1607  		}
  1608  		if len(actual) != 0 {
  1609  			t.Errorf("%s(%s, %q) returned non-empty value", name, report(tc.in), tc.arg)
  1610  		} else {
  1611  			actualNil := actual == nil
  1612  			outNil := tc.out == nil
  1613  			if actualNil != outNil {
  1614  				t.Errorf("%s(%s, %q) got nil %t; want nil %t", name, report(tc.in), tc.arg, actualNil, outNil)
  1615  			}
  1616  		}
  1617  	}
  1618  }
  1619  
  1620  type predicate struct {
  1621  	f    func(r rune) bool
  1622  	name string
  1623  }
  1624  
  1625  var isSpace = predicate{unicode.IsSpace, "IsSpace"}
  1626  var isDigit = predicate{unicode.IsDigit, "IsDigit"}
  1627  var isUpper = predicate{unicode.IsUpper, "IsUpper"}
  1628  var isValidRune = predicate{
  1629  	func(r rune) bool {
  1630  		return r != utf8.RuneError
  1631  	},
  1632  	"IsValidRune",
  1633  }
  1634  
  1635  type TrimFuncTest struct {
  1636  	f        predicate
  1637  	in       string
  1638  	trimOut  []byte
  1639  	leftOut  []byte
  1640  	rightOut []byte
  1641  }
  1642  
  1643  func not(p predicate) predicate {
  1644  	return predicate{
  1645  		func(r rune) bool {
  1646  			return !p.f(r)
  1647  		},
  1648  		"not " + p.name,
  1649  	}
  1650  }
  1651  
  1652  var trimFuncTests = []TrimFuncTest{
  1653  	{isSpace, space + " hello " + space,
  1654  		[]byte("hello"),
  1655  		[]byte("hello " + space),
  1656  		[]byte(space + " hello")},
  1657  	{isDigit, "\u0e50\u0e5212hello34\u0e50\u0e51",
  1658  		[]byte("hello"),
  1659  		[]byte("hello34\u0e50\u0e51"),
  1660  		[]byte("\u0e50\u0e5212hello")},
  1661  	{isUpper, "\u2C6F\u2C6F\u2C6F\u2C6FABCDhelloEF\u2C6F\u2C6FGH\u2C6F\u2C6F",
  1662  		[]byte("hello"),
  1663  		[]byte("helloEF\u2C6F\u2C6FGH\u2C6F\u2C6F"),
  1664  		[]byte("\u2C6F\u2C6F\u2C6F\u2C6FABCDhello")},
  1665  	{not(isSpace), "hello" + space + "hello",
  1666  		[]byte(space),
  1667  		[]byte(space + "hello"),
  1668  		[]byte("hello" + space)},
  1669  	{not(isDigit), "hello\u0e50\u0e521234\u0e50\u0e51helo",
  1670  		[]byte("\u0e50\u0e521234\u0e50\u0e51"),
  1671  		[]byte("\u0e50\u0e521234\u0e50\u0e51helo"),
  1672  		[]byte("hello\u0e50\u0e521234\u0e50\u0e51")},
  1673  	{isValidRune, "ab\xc0a\xc0cd",
  1674  		[]byte("\xc0a\xc0"),
  1675  		[]byte("\xc0a\xc0cd"),
  1676  		[]byte("ab\xc0a\xc0")},
  1677  	{not(isValidRune), "\xc0a\xc0",
  1678  		[]byte("a"),
  1679  		[]byte("a\xc0"),
  1680  		[]byte("\xc0a")},
  1681  	// The nils returned by TrimLeftFunc are odd behavior, but we need
  1682  	// to preserve backwards compatibility.
  1683  	{isSpace, "",
  1684  		nil,
  1685  		nil,
  1686  		[]byte("")},
  1687  	{isSpace, " ",
  1688  		nil,
  1689  		nil,
  1690  		[]byte("")},
  1691  }
  1692  
  1693  func TestTrimFunc(t *testing.T) {
  1694  	for _, tc := range trimFuncTests {
  1695  		trimmers := []struct {
  1696  			name string
  1697  			trim func(s []byte, f func(r rune) bool) []byte
  1698  			out  []byte
  1699  		}{
  1700  			{"TrimFunc", TrimFunc, tc.trimOut},
  1701  			{"TrimLeftFunc", TrimLeftFunc, tc.leftOut},
  1702  			{"TrimRightFunc", TrimRightFunc, tc.rightOut},
  1703  		}
  1704  		for _, trimmer := range trimmers {
  1705  			actual := trimmer.trim([]byte(tc.in), tc.f.f)
  1706  			if actual == nil && trimmer.out != nil {
  1707  				t.Errorf("%s(%q, %q) = nil; want %q", trimmer.name, tc.in, tc.f.name, trimmer.out)
  1708  			}
  1709  			if actual != nil && trimmer.out == nil {
  1710  				t.Errorf("%s(%q, %q) = %q; want nil", trimmer.name, tc.in, tc.f.name, actual)
  1711  			}
  1712  			if !Equal(actual, trimmer.out) {
  1713  				t.Errorf("%s(%q, %q) = %q; want %q", trimmer.name, tc.in, tc.f.name, actual, trimmer.out)
  1714  			}
  1715  		}
  1716  	}
  1717  }
  1718  
  1719  type IndexFuncTest struct {
  1720  	in          string
  1721  	f           predicate
  1722  	first, last int
  1723  }
  1724  
  1725  var indexFuncTests = []IndexFuncTest{
  1726  	{"", isValidRune, -1, -1},
  1727  	{"abc", isDigit, -1, -1},
  1728  	{"0123", isDigit, 0, 3},
  1729  	{"a1b", isDigit, 1, 1},
  1730  	{space, isSpace, 0, len(space) - 3}, // last rune in space is 3 bytes
  1731  	{"\u0e50\u0e5212hello34\u0e50\u0e51", isDigit, 0, 18},
  1732  	{"\u2C6F\u2C6F\u2C6F\u2C6FABCDhelloEF\u2C6F\u2C6FGH\u2C6F\u2C6F", isUpper, 0, 34},
  1733  	{"12\u0e50\u0e52hello34\u0e50\u0e51", not(isDigit), 8, 12},
  1734  
  1735  	// tests of invalid UTF-8
  1736  	{"\x801", isDigit, 1, 1},
  1737  	{"\x80abc", isDigit, -1, -1},
  1738  	{"\xc0a\xc0", isValidRune, 1, 1},
  1739  	{"\xc0a\xc0", not(isValidRune), 0, 2},
  1740  	{"\xc0☺\xc0", not(isValidRune), 0, 4},
  1741  	{"\xc0☺\xc0\xc0", not(isValidRune), 0, 5},
  1742  	{"ab\xc0a\xc0cd", not(isValidRune), 2, 4},
  1743  	{"a\xe0\x80cd", not(isValidRune), 1, 2},
  1744  }
  1745  
  1746  func TestIndexFunc(t *testing.T) {
  1747  	for _, tc := range indexFuncTests {
  1748  		first := IndexFunc([]byte(tc.in), tc.f.f)
  1749  		if first != tc.first {
  1750  			t.Errorf("IndexFunc(%q, %s) = %d; want %d", tc.in, tc.f.name, first, tc.first)
  1751  		}
  1752  		last := LastIndexFunc([]byte(tc.in), tc.f.f)
  1753  		if last != tc.last {
  1754  			t.Errorf("LastIndexFunc(%q, %s) = %d; want %d", tc.in, tc.f.name, last, tc.last)
  1755  		}
  1756  	}
  1757  }
  1758  
  1759  type ReplaceTest struct {
  1760  	in       string
  1761  	old, new string
  1762  	n        int
  1763  	out      string
  1764  }
  1765  
  1766  var ReplaceTests = []ReplaceTest{
  1767  	{"hello", "l", "L", 0, "hello"},
  1768  	{"hello", "l", "L", -1, "heLLo"},
  1769  	{"hello", "x", "X", -1, "hello"},
  1770  	{"", "x", "X", -1, ""},
  1771  	{"radar", "r", "<r>", -1, "<r>ada<r>"},
  1772  	{"", "", "<>", -1, "<>"},
  1773  	{"banana", "a", "<>", -1, "b<>n<>n<>"},
  1774  	{"banana", "a", "<>", 1, "b<>nana"},
  1775  	{"banana", "a", "<>", 1000, "b<>n<>n<>"},
  1776  	{"banana", "an", "<>", -1, "b<><>a"},
  1777  	{"banana", "ana", "<>", -1, "b<>na"},
  1778  	{"banana", "", "<>", -1, "<>b<>a<>n<>a<>n<>a<>"},
  1779  	{"banana", "", "<>", 10, "<>b<>a<>n<>a<>n<>a<>"},
  1780  	{"banana", "", "<>", 6, "<>b<>a<>n<>a<>n<>a"},
  1781  	{"banana", "", "<>", 5, "<>b<>a<>n<>a<>na"},
  1782  	{"banana", "", "<>", 1, "<>banana"},
  1783  	{"banana", "a", "a", -1, "banana"},
  1784  	{"banana", "a", "a", 1, "banana"},
  1785  	{"☺☻☹", "", "<>", -1, "<>☺<>☻<>☹<>"},
  1786  }
  1787  
  1788  func TestReplace(t *testing.T) {
  1789  	for _, tt := range ReplaceTests {
  1790  		var (
  1791  			in  = []byte(tt.in)
  1792  			old = []byte(tt.old)
  1793  			new = []byte(tt.new)
  1794  		)
  1795  		if !asan.Enabled {
  1796  			allocs := testing.AllocsPerRun(10, func() { Replace(in, old, new, tt.n) })
  1797  			if allocs > 1 {
  1798  				t.Errorf("Replace(%q, %q, %q, %d) allocates %.2f objects", tt.in, tt.old, tt.new, tt.n, allocs)
  1799  			}
  1800  		}
  1801  		in = append(in, "<spare>"...)
  1802  		in = in[:len(tt.in)]
  1803  		out := Replace(in, old, new, tt.n)
  1804  		if s := string(out); s != tt.out {
  1805  			t.Errorf("Replace(%q, %q, %q, %d) = %q, want %q", tt.in, tt.old, tt.new, tt.n, s, tt.out)
  1806  		}
  1807  		if cap(in) == cap(out) && &in[:1][0] == &out[:1][0] {
  1808  			t.Errorf("Replace(%q, %q, %q, %d) didn't copy", tt.in, tt.old, tt.new, tt.n)
  1809  		}
  1810  		if tt.n == -1 {
  1811  			out := ReplaceAll(in, old, new)
  1812  			if s := string(out); s != tt.out {
  1813  				t.Errorf("ReplaceAll(%q, %q, %q) = %q, want %q", tt.in, tt.old, tt.new, s, tt.out)
  1814  			}
  1815  		}
  1816  	}
  1817  }
  1818  
  1819  func FuzzReplace(f *testing.F) {
  1820  	for _, tt := range ReplaceTests {
  1821  		f.Add([]byte(tt.in), []byte(tt.old), []byte(tt.new), tt.n)
  1822  	}
  1823  	f.Fuzz(func(t *testing.T, in, old, new []byte, n int) {
  1824  		differentImpl := func(in, old, new []byte, n int) []byte {
  1825  			var out Buffer
  1826  			if n < 0 {
  1827  				n = math.MaxInt
  1828  			}
  1829  			for i := 0; i < len(in); {
  1830  				if n == 0 {
  1831  					out.Write(in[i:])
  1832  					break
  1833  				}
  1834  				if HasPrefix(in[i:], old) {
  1835  					out.Write(new)
  1836  					i += len(old)
  1837  					n--
  1838  					if len(old) != 0 {
  1839  						continue
  1840  					}
  1841  					if i == len(in) {
  1842  						break
  1843  					}
  1844  				}
  1845  				if len(old) == 0 {
  1846  					_, length := utf8.DecodeRune(in[i:])
  1847  					out.Write(in[i : i+length])
  1848  					i += length
  1849  				} else {
  1850  					out.WriteByte(in[i])
  1851  					i++
  1852  				}
  1853  			}
  1854  			if len(old) == 0 && n != 0 {
  1855  				out.Write(new)
  1856  			}
  1857  			return out.Bytes()
  1858  		}
  1859  		if simple, replace := differentImpl(in, old, new, n), Replace(in, old, new, n); !slices.Equal(simple, replace) {
  1860  			t.Errorf("The two implementations do not match %q != %q for Replace(%q, %q, %q, %d)", simple, replace, in, old, new, n)
  1861  		}
  1862  	})
  1863  }
  1864  
  1865  func BenchmarkReplace(b *testing.B) {
  1866  	for _, tt := range ReplaceTests {
  1867  		desc := fmt.Sprintf("%q %q %q %d", tt.in, tt.old, tt.new, tt.n)
  1868  		var (
  1869  			in  = []byte(tt.in)
  1870  			old = []byte(tt.old)
  1871  			new = []byte(tt.new)
  1872  		)
  1873  		b.Run(desc, func(b *testing.B) {
  1874  			b.ReportAllocs()
  1875  			for b.Loop() {
  1876  				Replace(in, old, new, tt.n)
  1877  			}
  1878  		})
  1879  	}
  1880  }
  1881  
  1882  type TitleTest struct {
  1883  	in, out string
  1884  }
  1885  
  1886  var TitleTests = []TitleTest{
  1887  	{"", ""},
  1888  	{"a", "A"},
  1889  	{" aaa aaa aaa ", " Aaa Aaa Aaa "},
  1890  	{" Aaa Aaa Aaa ", " Aaa Aaa Aaa "},
  1891  	{"123a456", "123a456"},
  1892  	{"double-blind", "Double-Blind"},
  1893  	{"ÿøû", "Ÿøû"},
  1894  	{"with_underscore", "With_underscore"},
  1895  	{"unicode \xe2\x80\xa8 line separator", "Unicode \xe2\x80\xa8 Line Separator"},
  1896  }
  1897  
  1898  func TestTitle(t *testing.T) {
  1899  	for _, tt := range TitleTests {
  1900  		if s := string(Title([]byte(tt.in))); s != tt.out {
  1901  			t.Errorf("Title(%q) = %q, want %q", tt.in, s, tt.out)
  1902  		}
  1903  	}
  1904  }
  1905  
  1906  var ToTitleTests = []TitleTest{
  1907  	{"", ""},
  1908  	{"a", "A"},
  1909  	{" aaa aaa aaa ", " AAA AAA AAA "},
  1910  	{" Aaa Aaa Aaa ", " AAA AAA AAA "},
  1911  	{"123a456", "123A456"},
  1912  	{"double-blind", "DOUBLE-BLIND"},
  1913  	{"ÿøû", "ŸØÛ"},
  1914  }
  1915  
  1916  func TestToTitle(t *testing.T) {
  1917  	for _, tt := range ToTitleTests {
  1918  		if s := string(ToTitle([]byte(tt.in))); s != tt.out {
  1919  			t.Errorf("ToTitle(%q) = %q, want %q", tt.in, s, tt.out)
  1920  		}
  1921  	}
  1922  }
  1923  
  1924  var EqualFoldTests = []struct {
  1925  	s, t string
  1926  	out  bool
  1927  }{
  1928  	{"abc", "abc", true},
  1929  	{"ABcd", "ABcd", true},
  1930  	{"123abc", "123ABC", true},
  1931  	{"αβδ", "ΑΒΔ", true},
  1932  	{"abc", "xyz", false},
  1933  	{"abc", "XYZ", false},
  1934  	{"abcdefghijk", "abcdefghijX", false},
  1935  	{"abcdefghijk", "abcdefghij\u212A", true},
  1936  	{"abcdefghijK", "abcdefghij\u212A", true},
  1937  	{"abcdefghijkz", "abcdefghij\u212Ay", false},
  1938  	{"abcdefghijKz", "abcdefghij\u212Ay", false},
  1939  }
  1940  
  1941  func TestEqualFold(t *testing.T) {
  1942  	for _, tt := range EqualFoldTests {
  1943  		if out := EqualFold([]byte(tt.s), []byte(tt.t)); out != tt.out {
  1944  			t.Errorf("EqualFold(%#q, %#q) = %v, want %v", tt.s, tt.t, out, tt.out)
  1945  		}
  1946  		if out := EqualFold([]byte(tt.t), []byte(tt.s)); out != tt.out {
  1947  			t.Errorf("EqualFold(%#q, %#q) = %v, want %v", tt.t, tt.s, out, tt.out)
  1948  		}
  1949  	}
  1950  }
  1951  
  1952  var cutTests = []struct {
  1953  	s, sep        string
  1954  	before, after string
  1955  	found         bool
  1956  }{
  1957  	{"abc", "b", "a", "c", true},
  1958  	{"abc", "a", "", "bc", true},
  1959  	{"abc", "c", "ab", "", true},
  1960  	{"abc", "abc", "", "", true},
  1961  	{"abc", "", "", "abc", true},
  1962  	{"abc", "d", "abc", "", false},
  1963  	{"", "d", "", "", false},
  1964  	{"", "", "", "", true},
  1965  }
  1966  
  1967  func TestCut(t *testing.T) {
  1968  	for _, tt := range cutTests {
  1969  		if before, after, found := Cut([]byte(tt.s), []byte(tt.sep)); string(before) != tt.before || string(after) != tt.after || found != tt.found {
  1970  			t.Errorf("Cut(%q, %q) = %q, %q, %v, want %q, %q, %v", tt.s, tt.sep, before, after, found, tt.before, tt.after, tt.found)
  1971  		}
  1972  	}
  1973  }
  1974  
  1975  var cutPrefixTests = []struct {
  1976  	s, sep string
  1977  	after  string
  1978  	found  bool
  1979  }{
  1980  	{"abc", "a", "bc", true},
  1981  	{"abc", "abc", "", true},
  1982  	{"abc", "", "abc", true},
  1983  	{"abc", "d", "abc", false},
  1984  	{"", "d", "", false},
  1985  	{"", "", "", true},
  1986  }
  1987  
  1988  func TestCutPrefix(t *testing.T) {
  1989  	for _, tt := range cutPrefixTests {
  1990  		if after, found := CutPrefix([]byte(tt.s), []byte(tt.sep)); string(after) != tt.after || found != tt.found {
  1991  			t.Errorf("CutPrefix(%q, %q) = %q, %v, want %q, %v", tt.s, tt.sep, after, found, tt.after, tt.found)
  1992  		}
  1993  	}
  1994  }
  1995  
  1996  var cutSuffixTests = []struct {
  1997  	s, sep string
  1998  	before string
  1999  	found  bool
  2000  }{
  2001  	{"abc", "bc", "a", true},
  2002  	{"abc", "abc", "", true},
  2003  	{"abc", "", "abc", true},
  2004  	{"abc", "d", "abc", false},
  2005  	{"", "d", "", false},
  2006  	{"", "", "", true},
  2007  }
  2008  
  2009  func TestCutSuffix(t *testing.T) {
  2010  	for _, tt := range cutSuffixTests {
  2011  		if before, found := CutSuffix([]byte(tt.s), []byte(tt.sep)); string(before) != tt.before || found != tt.found {
  2012  			t.Errorf("CutSuffix(%q, %q) = %q, %v, want %q, %v", tt.s, tt.sep, before, found, tt.before, tt.found)
  2013  		}
  2014  	}
  2015  }
  2016  
  2017  func TestBufferGrowNegative(t *testing.T) {
  2018  	defer func() {
  2019  		if err := recover(); err == nil {
  2020  			t.Fatal("Grow(-1) should have panicked")
  2021  		}
  2022  	}()
  2023  	var b Buffer
  2024  	b.Grow(-1)
  2025  }
  2026  
  2027  func TestBufferTruncateNegative(t *testing.T) {
  2028  	defer func() {
  2029  		if err := recover(); err == nil {
  2030  			t.Fatal("Truncate(-1) should have panicked")
  2031  		}
  2032  	}()
  2033  	var b Buffer
  2034  	b.Truncate(-1)
  2035  }
  2036  
  2037  func TestBufferTruncateOutOfRange(t *testing.T) {
  2038  	defer func() {
  2039  		if err := recover(); err == nil {
  2040  			t.Fatal("Truncate(20) should have panicked")
  2041  		}
  2042  	}()
  2043  	var b Buffer
  2044  	b.Write(make([]byte, 10))
  2045  	b.Truncate(20)
  2046  }
  2047  
  2048  var containsTests = []struct {
  2049  	b, subslice []byte
  2050  	want        bool
  2051  }{
  2052  	{[]byte("hello"), []byte("hel"), true},
  2053  	{[]byte("日本語"), []byte("日本"), true},
  2054  	{[]byte("hello"), []byte("Hello, world"), false},
  2055  	{[]byte("東京"), []byte("京東"), false},
  2056  }
  2057  
  2058  func TestContains(t *testing.T) {
  2059  	for _, tt := range containsTests {
  2060  		if got := Contains(tt.b, tt.subslice); got != tt.want {
  2061  			t.Errorf("Contains(%q, %q) = %v, want %v", tt.b, tt.subslice, got, tt.want)
  2062  		}
  2063  	}
  2064  }
  2065  
  2066  var ContainsAnyTests = []struct {
  2067  	b        []byte
  2068  	substr   string
  2069  	expected bool
  2070  }{
  2071  	{[]byte(""), "", false},
  2072  	{[]byte(""), "a", false},
  2073  	{[]byte(""), "abc", false},
  2074  	{[]byte("a"), "", false},
  2075  	{[]byte("a"), "a", true},
  2076  	{[]byte("aaa"), "a", true},
  2077  	{[]byte("abc"), "xyz", false},
  2078  	{[]byte("abc"), "xcz", true},
  2079  	{[]byte("a☺b☻c☹d"), "uvw☻xyz", true},
  2080  	{[]byte("aRegExp*"), ".(|)*+?^$[]", true},
  2081  	{[]byte(dots + dots + dots), " ", false},
  2082  }
  2083  
  2084  func TestContainsAny(t *testing.T) {
  2085  	for _, ct := range ContainsAnyTests {
  2086  		if ContainsAny(ct.b, ct.substr) != ct.expected {
  2087  			t.Errorf("ContainsAny(%s, %s) = %v, want %v",
  2088  				ct.b, ct.substr, !ct.expected, ct.expected)
  2089  		}
  2090  	}
  2091  }
  2092  
  2093  var ContainsRuneTests = []struct {
  2094  	b        []byte
  2095  	r        rune
  2096  	expected bool
  2097  }{
  2098  	{[]byte(""), 'a', false},
  2099  	{[]byte("a"), 'a', true},
  2100  	{[]byte("aaa"), 'a', true},
  2101  	{[]byte("abc"), 'y', false},
  2102  	{[]byte("abc"), 'c', true},
  2103  	{[]byte("a☺b☻c☹d"), 'x', false},
  2104  	{[]byte("a☺b☻c☹d"), '☻', true},
  2105  	{[]byte("aRegExp*"), '*', true},
  2106  }
  2107  
  2108  func TestContainsRune(t *testing.T) {
  2109  	for _, ct := range ContainsRuneTests {
  2110  		if ContainsRune(ct.b, ct.r) != ct.expected {
  2111  			t.Errorf("ContainsRune(%q, %q) = %v, want %v",
  2112  				ct.b, ct.r, !ct.expected, ct.expected)
  2113  		}
  2114  	}
  2115  }
  2116  
  2117  func TestContainsFunc(t *testing.T) {
  2118  	for _, ct := range ContainsRuneTests {
  2119  		if ContainsFunc(ct.b, func(r rune) bool {
  2120  			return ct.r == r
  2121  		}) != ct.expected {
  2122  			t.Errorf("ContainsFunc(%q, func(%q)) = %v, want %v",
  2123  				ct.b, ct.r, !ct.expected, ct.expected)
  2124  		}
  2125  	}
  2126  }
  2127  
  2128  var makeFieldsInput = func() []byte {
  2129  	x := make([]byte, 1<<20)
  2130  	// Input is ~10% space, ~10% 2-byte UTF-8, rest ASCII non-space.
  2131  	for i := range x {
  2132  		switch rand.Intn(10) {
  2133  		case 0:
  2134  			x[i] = ' '
  2135  		case 1:
  2136  			if i > 0 && x[i-1] == 'x' {
  2137  				copy(x[i-1:], "χ")
  2138  				break
  2139  			}
  2140  			fallthrough
  2141  		default:
  2142  			x[i] = 'x'
  2143  		}
  2144  	}
  2145  	return x
  2146  }
  2147  
  2148  var makeFieldsInputASCII = func() []byte {
  2149  	x := make([]byte, 1<<20)
  2150  	// Input is ~10% space, rest ASCII non-space.
  2151  	for i := range x {
  2152  		if rand.Intn(10) == 0 {
  2153  			x[i] = ' '
  2154  		} else {
  2155  			x[i] = 'x'
  2156  		}
  2157  	}
  2158  	return x
  2159  }
  2160  
  2161  var bytesdata = []struct {
  2162  	name string
  2163  	data []byte
  2164  }{
  2165  	{"ASCII", makeFieldsInputASCII()},
  2166  	{"Mixed", makeFieldsInput()},
  2167  }
  2168  
  2169  func BenchmarkFields(b *testing.B) {
  2170  	for _, sd := range bytesdata {
  2171  		b.Run(sd.name, func(b *testing.B) {
  2172  			for j := 1 << 4; j <= 1<<20; j <<= 4 {
  2173  				b.Run(fmt.Sprintf("%d", j), func(b *testing.B) {
  2174  					b.ReportAllocs()
  2175  					b.SetBytes(int64(j))
  2176  					data := sd.data[:j]
  2177  					for i := 0; i < b.N; i++ {
  2178  						Fields(data)
  2179  					}
  2180  				})
  2181  			}
  2182  		})
  2183  	}
  2184  }
  2185  
  2186  func BenchmarkFieldsFunc(b *testing.B) {
  2187  	for _, sd := range bytesdata {
  2188  		b.Run(sd.name, func(b *testing.B) {
  2189  			for j := 1 << 4; j <= 1<<20; j <<= 4 {
  2190  				b.Run(fmt.Sprintf("%d", j), func(b *testing.B) {
  2191  					b.ReportAllocs()
  2192  					b.SetBytes(int64(j))
  2193  					data := sd.data[:j]
  2194  					for i := 0; i < b.N; i++ {
  2195  						FieldsFunc(data, unicode.IsSpace)
  2196  					}
  2197  				})
  2198  			}
  2199  		})
  2200  	}
  2201  }
  2202  
  2203  func BenchmarkTrimSpace(b *testing.B) {
  2204  	tests := []struct {
  2205  		name  string
  2206  		input []byte
  2207  	}{
  2208  		{"NoTrim", []byte("typical")},
  2209  		{"ASCII", []byte("  foo bar  ")},
  2210  		{"SomeNonASCII", []byte("    \u2000\t\r\n x\t\t\r\r\ny\n \u3000    ")},
  2211  		{"JustNonASCII", []byte("\u2000\u2000\u2000☺☺☺☺\u3000\u3000\u3000")},
  2212  	}
  2213  	for _, test := range tests {
  2214  		b.Run(test.name, func(b *testing.B) {
  2215  			for i := 0; i < b.N; i++ {
  2216  				TrimSpace(test.input)
  2217  			}
  2218  		})
  2219  	}
  2220  }
  2221  
  2222  func BenchmarkToValidUTF8(b *testing.B) {
  2223  	tests := []struct {
  2224  		name  string
  2225  		input []byte
  2226  	}{
  2227  		{"Valid", []byte("typical")},
  2228  		{"InvalidASCII", []byte("foo\xffbar")},
  2229  		{"InvalidNonASCII", []byte("日本語\xff日本語")},
  2230  	}
  2231  	replacement := []byte("\uFFFD")
  2232  	b.ResetTimer()
  2233  	for _, test := range tests {
  2234  		b.Run(test.name, func(b *testing.B) {
  2235  			for i := 0; i < b.N; i++ {
  2236  				ToValidUTF8(test.input, replacement)
  2237  			}
  2238  		})
  2239  	}
  2240  }
  2241  
  2242  func makeBenchInputHard() []byte {
  2243  	tokens := [...]string{
  2244  		"<a>", "<p>", "<b>", "<strong>",
  2245  		"</a>", "</p>", "</b>", "</strong>",
  2246  		"hello", "world",
  2247  	}
  2248  	x := make([]byte, 0, 1<<20)
  2249  	for {
  2250  		i := rand.Intn(len(tokens))
  2251  		if len(x)+len(tokens[i]) >= 1<<20 {
  2252  			break
  2253  		}
  2254  		x = append(x, tokens[i]...)
  2255  	}
  2256  	return x
  2257  }
  2258  
  2259  var benchInputHard = makeBenchInputHard()
  2260  
  2261  func benchmarkIndexHard(b *testing.B, sep []byte) {
  2262  	n := Index(benchInputHard, sep)
  2263  	if n < 0 {
  2264  		n = len(benchInputHard)
  2265  	}
  2266  	b.SetBytes(int64(n))
  2267  	for i := 0; i < b.N; i++ {
  2268  		Index(benchInputHard, sep)
  2269  	}
  2270  }
  2271  
  2272  func benchmarkLastIndexHard(b *testing.B, sep []byte) {
  2273  	for i := 0; i < b.N; i++ {
  2274  		LastIndex(benchInputHard, sep)
  2275  	}
  2276  }
  2277  
  2278  func benchmarkCountHard(b *testing.B, sep []byte) {
  2279  	for i := 0; i < b.N; i++ {
  2280  		Count(benchInputHard, sep)
  2281  	}
  2282  }
  2283  
  2284  func BenchmarkIndexHard1(b *testing.B) { benchmarkIndexHard(b, []byte("<>")) }
  2285  func BenchmarkIndexHard2(b *testing.B) { benchmarkIndexHard(b, []byte("</pre>")) }
  2286  func BenchmarkIndexHard3(b *testing.B) { benchmarkIndexHard(b, []byte("<b>hello world</b>")) }
  2287  func BenchmarkIndexHard4(b *testing.B) {
  2288  	benchmarkIndexHard(b, []byte("<pre><b>hello</b><strong>world</strong></pre>"))
  2289  }
  2290  
  2291  func BenchmarkLastIndexHard1(b *testing.B) { benchmarkLastIndexHard(b, []byte("<>")) }
  2292  func BenchmarkLastIndexHard2(b *testing.B) { benchmarkLastIndexHard(b, []byte("</pre>")) }
  2293  func BenchmarkLastIndexHard3(b *testing.B) { benchmarkLastIndexHard(b, []byte("<b>hello world</b>")) }
  2294  
  2295  func BenchmarkCountHard1(b *testing.B) { benchmarkCountHard(b, []byte("<>")) }
  2296  func BenchmarkCountHard2(b *testing.B) { benchmarkCountHard(b, []byte("</pre>")) }
  2297  func BenchmarkCountHard3(b *testing.B) { benchmarkCountHard(b, []byte("<b>hello world</b>")) }
  2298  
  2299  func BenchmarkSplitEmptySeparator(b *testing.B) {
  2300  	for i := 0; i < b.N; i++ {
  2301  		Split(benchInputHard, nil)
  2302  	}
  2303  }
  2304  
  2305  func BenchmarkSplitSingleByteSeparator(b *testing.B) {
  2306  	sep := []byte("/")
  2307  	for i := 0; i < b.N; i++ {
  2308  		Split(benchInputHard, sep)
  2309  	}
  2310  }
  2311  
  2312  func BenchmarkSplitMultiByteSeparator(b *testing.B) {
  2313  	sep := []byte("hello")
  2314  	for i := 0; i < b.N; i++ {
  2315  		Split(benchInputHard, sep)
  2316  	}
  2317  }
  2318  
  2319  func BenchmarkSplitNSingleByteSeparator(b *testing.B) {
  2320  	sep := []byte("/")
  2321  	for i := 0; i < b.N; i++ {
  2322  		SplitN(benchInputHard, sep, 10)
  2323  	}
  2324  }
  2325  
  2326  func BenchmarkSplitNMultiByteSeparator(b *testing.B) {
  2327  	sep := []byte("hello")
  2328  	for i := 0; i < b.N; i++ {
  2329  		SplitN(benchInputHard, sep, 10)
  2330  	}
  2331  }
  2332  
  2333  func BenchmarkRepeat(b *testing.B) {
  2334  	for i := 0; i < b.N; i++ {
  2335  		Repeat([]byte("-"), 80)
  2336  	}
  2337  }
  2338  
  2339  func BenchmarkRepeatLarge(b *testing.B) {
  2340  	s := Repeat([]byte("@"), 8*1024)
  2341  	for j := 8; j <= 30; j++ {
  2342  		for _, k := range []int{1, 16, 4097} {
  2343  			s := s[:k]
  2344  			n := (1 << j) / k
  2345  			if n == 0 {
  2346  				continue
  2347  			}
  2348  			b.Run(fmt.Sprintf("%d/%d", 1<<j, k), func(b *testing.B) {
  2349  				for i := 0; i < b.N; i++ {
  2350  					Repeat(s, n)
  2351  				}
  2352  				b.SetBytes(int64(n * len(s)))
  2353  			})
  2354  		}
  2355  	}
  2356  }
  2357  
  2358  func BenchmarkBytesCompare(b *testing.B) {
  2359  	for n := 1; n <= 2048; n <<= 1 {
  2360  		b.Run(fmt.Sprint(n), func(b *testing.B) {
  2361  			var x = make([]byte, n)
  2362  			var y = make([]byte, n)
  2363  
  2364  			for i := 0; i < n; i++ {
  2365  				x[i] = 'a'
  2366  			}
  2367  
  2368  			for i := 0; i < n; i++ {
  2369  				y[i] = 'a'
  2370  			}
  2371  
  2372  			b.ResetTimer()
  2373  			for i := 0; i < b.N; i++ {
  2374  				Compare(x, y)
  2375  			}
  2376  		})
  2377  	}
  2378  }
  2379  
  2380  func BenchmarkIndexAnyASCII(b *testing.B) {
  2381  	x := Repeat([]byte{'#'}, 2048) // Never matches set
  2382  	cs := "0123456789abcdefghijklmnopqrstuvwxyz0123456789abcdefghijklmnopqrstuvwxyz"
  2383  	for k := 1; k <= 2048; k <<= 4 {
  2384  		for j := 1; j <= 64; j <<= 1 {
  2385  			b.Run(fmt.Sprintf("%d:%d", k, j), func(b *testing.B) {
  2386  				for i := 0; i < b.N; i++ {
  2387  					IndexAny(x[:k], cs[:j])
  2388  				}
  2389  			})
  2390  		}
  2391  	}
  2392  }
  2393  
  2394  func BenchmarkIndexAnyUTF8(b *testing.B) {
  2395  	x := Repeat([]byte{'#'}, 2048) // Never matches set
  2396  	cs := "你好世界, hello world. 你好世界, hello world. 你好世界, hello world."
  2397  	for k := 1; k <= 2048; k <<= 4 {
  2398  		for j := 1; j <= 64; j <<= 1 {
  2399  			b.Run(fmt.Sprintf("%d:%d", k, j), func(b *testing.B) {
  2400  				for i := 0; i < b.N; i++ {
  2401  					IndexAny(x[:k], cs[:j])
  2402  				}
  2403  			})
  2404  		}
  2405  	}
  2406  }
  2407  
  2408  func BenchmarkLastIndexAnyASCII(b *testing.B) {
  2409  	x := Repeat([]byte{'#'}, 2048) // Never matches set
  2410  	cs := "0123456789abcdefghijklmnopqrstuvwxyz0123456789abcdefghijklmnopqrstuvwxyz"
  2411  	for k := 1; k <= 2048; k <<= 4 {
  2412  		for j := 1; j <= 64; j <<= 1 {
  2413  			b.Run(fmt.Sprintf("%d:%d", k, j), func(b *testing.B) {
  2414  				for i := 0; i < b.N; i++ {
  2415  					LastIndexAny(x[:k], cs[:j])
  2416  				}
  2417  			})
  2418  		}
  2419  	}
  2420  }
  2421  
  2422  func BenchmarkLastIndexAnyUTF8(b *testing.B) {
  2423  	x := Repeat([]byte{'#'}, 2048) // Never matches set
  2424  	cs := "你好世界, hello world. 你好世界, hello world. 你好世界, hello world."
  2425  	for k := 1; k <= 2048; k <<= 4 {
  2426  		for j := 1; j <= 64; j <<= 1 {
  2427  			b.Run(fmt.Sprintf("%d:%d", k, j), func(b *testing.B) {
  2428  				for i := 0; i < b.N; i++ {
  2429  					LastIndexAny(x[:k], cs[:j])
  2430  				}
  2431  			})
  2432  		}
  2433  	}
  2434  }
  2435  
  2436  func BenchmarkTrimASCII(b *testing.B) {
  2437  	cs := "0123456789abcdef"
  2438  	for k := 1; k <= 4096; k <<= 4 {
  2439  		for j := 1; j <= 16; j <<= 1 {
  2440  			b.Run(fmt.Sprintf("%d:%d", k, j), func(b *testing.B) {
  2441  				x := Repeat([]byte(cs[:j]), k) // Always matches set
  2442  				for i := 0; i < b.N; i++ {
  2443  					Trim(x[:k], cs[:j])
  2444  				}
  2445  			})
  2446  		}
  2447  	}
  2448  }
  2449  
  2450  func BenchmarkTrimByte(b *testing.B) {
  2451  	x := []byte("  the quick brown fox   ")
  2452  	for i := 0; i < b.N; i++ {
  2453  		Trim(x, " ")
  2454  	}
  2455  }
  2456  
  2457  func BenchmarkIndexPeriodic(b *testing.B) {
  2458  	key := []byte{1, 1}
  2459  	for _, skip := range [...]int{2, 4, 8, 16, 32, 64} {
  2460  		b.Run(fmt.Sprintf("IndexPeriodic%d", skip), func(b *testing.B) {
  2461  			buf := make([]byte, 1<<16)
  2462  			for i := 0; i < len(buf); i += skip {
  2463  				buf[i] = 1
  2464  			}
  2465  			for i := 0; i < b.N; i++ {
  2466  				Index(buf, key)
  2467  			}
  2468  		})
  2469  	}
  2470  }
  2471  
  2472  func TestClone(t *testing.T) {
  2473  	var cloneTests = [][]byte{
  2474  		[]byte(nil),
  2475  		[]byte{},
  2476  		Clone([]byte{}),
  2477  		[]byte(strings.Repeat("a", 42))[:0],
  2478  		[]byte(strings.Repeat("a", 42))[:0:0],
  2479  		[]byte("short"),
  2480  		[]byte(strings.Repeat("a", 42)),
  2481  	}
  2482  	for _, input := range cloneTests {
  2483  		clone := Clone(input)
  2484  		if !Equal(clone, input) {
  2485  			t.Errorf("Clone(%q) = %q; want %q", input, clone, input)
  2486  		}
  2487  
  2488  		if input == nil && clone != nil {
  2489  			t.Errorf("Clone(%#v) return value should be equal to nil slice.", input)
  2490  		}
  2491  
  2492  		if input != nil && clone == nil {
  2493  			t.Errorf("Clone(%#v) return value should not be equal to nil slice.", input)
  2494  		}
  2495  
  2496  		if cap(input) != 0 && unsafe.SliceData(input) == unsafe.SliceData(clone) {
  2497  			t.Errorf("Clone(%q) return value should not reference inputs backing memory.", input)
  2498  		}
  2499  	}
  2500  }
  2501  

View as plain text