David K. Bainbridge | 215e024 | 2017-09-05 23:18:24 -0700 | [diff] [blame^] | 1 | // Copyright 2011 The Go Authors. All rights reserved. |
| 2 | // Use of this source code is governed by a BSD-style |
| 3 | // license that can be found in the LICENSE file. |
| 4 | |
| 5 | package norm |
| 6 | |
| 7 | import ( |
| 8 | "bufio" |
| 9 | "bytes" |
| 10 | "fmt" |
| 11 | "regexp" |
| 12 | "runtime" |
| 13 | "strconv" |
| 14 | "strings" |
| 15 | "sync" |
| 16 | "testing" |
| 17 | "time" |
| 18 | "unicode/utf8" |
| 19 | |
| 20 | "golang.org/x/text/internal/gen" |
| 21 | "golang.org/x/text/internal/testtext" |
| 22 | ) |
| 23 | |
| 24 | var once sync.Once |
| 25 | |
| 26 | func skipShort(t *testing.T) { |
| 27 | testtext.SkipIfNotLong(t) |
| 28 | |
| 29 | once.Do(func() { loadTestData(t) }) |
| 30 | } |
| 31 | |
| 32 | // This regression test runs the test set in NormalizationTest.txt |
| 33 | // (taken from http://www.unicode.org/Public/<unicode.Version>/ucd/). |
| 34 | // |
| 35 | // NormalizationTest.txt has form: |
| 36 | // @Part0 # Specific cases |
| 37 | // # |
| 38 | // 1E0A;1E0A;0044 0307;1E0A;0044 0307; # (Ḋ; Ḋ; D◌̇; Ḋ; D◌̇; ) LATIN CAPITAL LETTER D WITH DOT ABOVE |
| 39 | // 1E0C;1E0C;0044 0323;1E0C;0044 0323; # (Ḍ; Ḍ; D◌̣; Ḍ; D◌̣; ) LATIN CAPITAL LETTER D WITH DOT BELOW |
| 40 | // |
| 41 | // Each test has 5 columns (c1, c2, c3, c4, c5), where |
| 42 | // (c1, c2, c3, c4, c5) == (c1, NFC(c1), NFD(c1), NFKC(c1), NFKD(c1)) |
| 43 | // |
| 44 | // CONFORMANCE: |
| 45 | // 1. The following invariants must be true for all conformant implementations |
| 46 | // |
| 47 | // NFC |
| 48 | // c2 == NFC(c1) == NFC(c2) == NFC(c3) |
| 49 | // c4 == NFC(c4) == NFC(c5) |
| 50 | // |
| 51 | // NFD |
| 52 | // c3 == NFD(c1) == NFD(c2) == NFD(c3) |
| 53 | // c5 == NFD(c4) == NFD(c5) |
| 54 | // |
| 55 | // NFKC |
| 56 | // c4 == NFKC(c1) == NFKC(c2) == NFKC(c3) == NFKC(c4) == NFKC(c5) |
| 57 | // |
| 58 | // NFKD |
| 59 | // c5 == NFKD(c1) == NFKD(c2) == NFKD(c3) == NFKD(c4) == NFKD(c5) |
| 60 | // |
| 61 | // 2. For every code point X assigned in this version of Unicode that is not |
| 62 | // specifically listed in Part 1, the following invariants must be true |
| 63 | // for all conformant implementations: |
| 64 | // |
| 65 | // X == NFC(X) == NFD(X) == NFKC(X) == NFKD(X) |
| 66 | // |
| 67 | |
| 68 | // Column types. |
| 69 | const ( |
| 70 | cRaw = iota |
| 71 | cNFC |
| 72 | cNFD |
| 73 | cNFKC |
| 74 | cNFKD |
| 75 | cMaxColumns |
| 76 | ) |
| 77 | |
| 78 | // Holds data from NormalizationTest.txt |
| 79 | var part []Part |
| 80 | |
| 81 | type Part struct { |
| 82 | name string |
| 83 | number int |
| 84 | tests []Test |
| 85 | } |
| 86 | |
| 87 | type Test struct { |
| 88 | name string |
| 89 | partnr int |
| 90 | number int |
| 91 | r rune // used for character by character test |
| 92 | cols [cMaxColumns]string // Each has 5 entries, see below. |
| 93 | } |
| 94 | |
| 95 | func (t Test) Name() string { |
| 96 | if t.number < 0 { |
| 97 | return part[t.partnr].name |
| 98 | } |
| 99 | return fmt.Sprintf("%s:%d", part[t.partnr].name, t.number) |
| 100 | } |
| 101 | |
| 102 | var partRe = regexp.MustCompile(`@Part(\d) # (.*)$`) |
| 103 | var testRe = regexp.MustCompile(`^` + strings.Repeat(`([\dA-F ]+);`, 5) + ` # (.*)$`) |
| 104 | |
| 105 | var counter int |
| 106 | |
| 107 | // Load the data form NormalizationTest.txt |
| 108 | func loadTestData(t *testing.T) { |
| 109 | f := gen.OpenUCDFile("NormalizationTest.txt") |
| 110 | defer f.Close() |
| 111 | scanner := bufio.NewScanner(f) |
| 112 | for scanner.Scan() { |
| 113 | line := scanner.Text() |
| 114 | if len(line) == 0 || line[0] == '#' { |
| 115 | continue |
| 116 | } |
| 117 | m := partRe.FindStringSubmatch(line) |
| 118 | if m != nil { |
| 119 | if len(m) < 3 { |
| 120 | t.Fatal("Failed to parse Part: ", line) |
| 121 | } |
| 122 | i, err := strconv.Atoi(m[1]) |
| 123 | if err != nil { |
| 124 | t.Fatal(err) |
| 125 | } |
| 126 | name := m[2] |
| 127 | part = append(part, Part{name: name[:len(name)-1], number: i}) |
| 128 | continue |
| 129 | } |
| 130 | m = testRe.FindStringSubmatch(line) |
| 131 | if m == nil || len(m) < 7 { |
| 132 | t.Fatalf(`Failed to parse: "%s" result: %#v`, line, m) |
| 133 | } |
| 134 | test := Test{name: m[6], partnr: len(part) - 1, number: counter} |
| 135 | counter++ |
| 136 | for j := 1; j < len(m)-1; j++ { |
| 137 | for _, split := range strings.Split(m[j], " ") { |
| 138 | r, err := strconv.ParseUint(split, 16, 64) |
| 139 | if err != nil { |
| 140 | t.Fatal(err) |
| 141 | } |
| 142 | if test.r == 0 { |
| 143 | // save for CharacterByCharacterTests |
| 144 | test.r = rune(r) |
| 145 | } |
| 146 | var buf [utf8.UTFMax]byte |
| 147 | sz := utf8.EncodeRune(buf[:], rune(r)) |
| 148 | test.cols[j-1] += string(buf[:sz]) |
| 149 | } |
| 150 | } |
| 151 | part := &part[len(part)-1] |
| 152 | part.tests = append(part.tests, test) |
| 153 | } |
| 154 | if scanner.Err() != nil { |
| 155 | t.Fatal(scanner.Err()) |
| 156 | } |
| 157 | } |
| 158 | |
| 159 | func cmpResult(t *testing.T, tc *Test, name string, f Form, gold, test, result string) { |
| 160 | if gold != result { |
| 161 | t.Errorf("%s:%s: %s(%+q)=%+q; want %+q: %s", |
| 162 | tc.Name(), name, fstr[f], test, result, gold, tc.name) |
| 163 | } |
| 164 | } |
| 165 | |
| 166 | func cmpIsNormal(t *testing.T, tc *Test, name string, f Form, test string, result, want bool) { |
| 167 | if result != want { |
| 168 | t.Errorf("%s:%s: %s(%+q)=%v; want %v", tc.Name(), name, fstr[f], test, result, want) |
| 169 | } |
| 170 | } |
| 171 | |
| 172 | func doTest(t *testing.T, tc *Test, f Form, gold, test string) { |
| 173 | testb := []byte(test) |
| 174 | result := f.Bytes(testb) |
| 175 | cmpResult(t, tc, "Bytes", f, gold, test, string(result)) |
| 176 | |
| 177 | sresult := f.String(test) |
| 178 | cmpResult(t, tc, "String", f, gold, test, sresult) |
| 179 | |
| 180 | acc := []byte{} |
| 181 | i := Iter{} |
| 182 | i.InitString(f, test) |
| 183 | for !i.Done() { |
| 184 | acc = append(acc, i.Next()...) |
| 185 | } |
| 186 | cmpResult(t, tc, "Iter.Next", f, gold, test, string(acc)) |
| 187 | |
| 188 | buf := make([]byte, 128) |
| 189 | acc = nil |
| 190 | for p := 0; p < len(testb); { |
| 191 | nDst, nSrc, _ := f.Transform(buf, testb[p:], true) |
| 192 | acc = append(acc, buf[:nDst]...) |
| 193 | p += nSrc |
| 194 | } |
| 195 | cmpResult(t, tc, "Transform", f, gold, test, string(acc)) |
| 196 | |
| 197 | for i := range test { |
| 198 | out := f.Append(f.Bytes([]byte(test[:i])), []byte(test[i:])...) |
| 199 | cmpResult(t, tc, fmt.Sprintf(":Append:%d", i), f, gold, test, string(out)) |
| 200 | } |
| 201 | cmpIsNormal(t, tc, "IsNormal", f, test, f.IsNormal([]byte(test)), test == gold) |
| 202 | cmpIsNormal(t, tc, "IsNormalString", f, test, f.IsNormalString(test), test == gold) |
| 203 | } |
| 204 | |
| 205 | func doConformanceTests(t *testing.T, tc *Test, partn int) { |
| 206 | for i := 0; i <= 2; i++ { |
| 207 | doTest(t, tc, NFC, tc.cols[1], tc.cols[i]) |
| 208 | doTest(t, tc, NFD, tc.cols[2], tc.cols[i]) |
| 209 | doTest(t, tc, NFKC, tc.cols[3], tc.cols[i]) |
| 210 | doTest(t, tc, NFKD, tc.cols[4], tc.cols[i]) |
| 211 | } |
| 212 | for i := 3; i <= 4; i++ { |
| 213 | doTest(t, tc, NFC, tc.cols[3], tc.cols[i]) |
| 214 | doTest(t, tc, NFD, tc.cols[4], tc.cols[i]) |
| 215 | doTest(t, tc, NFKC, tc.cols[3], tc.cols[i]) |
| 216 | doTest(t, tc, NFKD, tc.cols[4], tc.cols[i]) |
| 217 | } |
| 218 | } |
| 219 | |
| 220 | func TestCharacterByCharacter(t *testing.T) { |
| 221 | skipShort(t) |
| 222 | tests := part[1].tests |
| 223 | var last rune = 0 |
| 224 | for i := 0; i <= len(tests); i++ { // last one is special case |
| 225 | var r rune |
| 226 | if i == len(tests) { |
| 227 | r = 0x2FA1E // Don't have to go to 0x10FFFF |
| 228 | } else { |
| 229 | r = tests[i].r |
| 230 | } |
| 231 | for last++; last < r; last++ { |
| 232 | // Check all characters that were not explicitly listed in the test. |
| 233 | tc := &Test{partnr: 1, number: -1} |
| 234 | char := string(last) |
| 235 | doTest(t, tc, NFC, char, char) |
| 236 | doTest(t, tc, NFD, char, char) |
| 237 | doTest(t, tc, NFKC, char, char) |
| 238 | doTest(t, tc, NFKD, char, char) |
| 239 | } |
| 240 | if i < len(tests) { |
| 241 | doConformanceTests(t, &tests[i], 1) |
| 242 | } |
| 243 | } |
| 244 | } |
| 245 | |
| 246 | func TestStandardTests(t *testing.T) { |
| 247 | skipShort(t) |
| 248 | for _, j := range []int{0, 2, 3} { |
| 249 | for _, test := range part[j].tests { |
| 250 | doConformanceTests(t, &test, j) |
| 251 | } |
| 252 | } |
| 253 | } |
| 254 | |
| 255 | // TestPerformance verifies that normalization is O(n). If any of the |
| 256 | // code does not properly check for maxCombiningChars, normalization |
| 257 | // may exhibit O(n**2) behavior. |
| 258 | func TestPerformance(t *testing.T) { |
| 259 | skipShort(t) |
| 260 | runtime.GOMAXPROCS(2) |
| 261 | success := make(chan bool, 1) |
| 262 | go func() { |
| 263 | buf := bytes.Repeat([]byte("\u035D"), 1024*1024) |
| 264 | buf = append(buf, "\u035B"...) |
| 265 | NFC.Append(nil, buf...) |
| 266 | success <- true |
| 267 | }() |
| 268 | timeout := time.After(1 * time.Second) |
| 269 | select { |
| 270 | case <-success: |
| 271 | // test completed before the timeout |
| 272 | case <-timeout: |
| 273 | t.Errorf(`unexpectedly long time to complete PerformanceTest`) |
| 274 | } |
| 275 | } |