|
| 1 | +using System.Globalization; |
| 2 | +using System.Text; |
| 3 | +using SqlServerSimulator.Storage; |
| 4 | +using static Microsoft.VisualStudio.TestTools.UnitTesting.Assert; |
| 5 | + |
| 6 | +namespace SqlServerSimulator; |
| 7 | + |
| 8 | +/// <summary> |
| 9 | +/// Guards the <c>IEqualityComparer<string></c> hash contract on |
| 10 | +/// <see cref="Collation.SqlLatin1Cp1CiAsCollation"/>: whenever |
| 11 | +/// <c>Equals(x, y)</c> is true — including across the repertoire boundary, |
| 12 | +/// where equality routes through the inner <see cref="CompareInfo"/> — |
| 13 | +/// <c>GetHashCode</c> must agree. The hybrid hashes in-repertoire strings |
| 14 | +/// by SQL sort-weight runs and everything else through canonicalization + |
| 15 | +/// the inner hash, so the boundary is where inconsistency would hide. |
| 16 | +/// Derivation notes live in <c>docs/claude/collations.md</c>. |
| 17 | +/// </summary> |
| 18 | +[TestClass] |
| 19 | +public sealed class CollationHashConsistencyTests |
| 20 | +{ |
| 21 | + private static readonly Collation Nvarchar = Collation.Get(Collation.SqlLatin1Cp1CiAsCollation.CollationName); |
| 22 | + |
| 23 | + private static readonly Collation Varchar = Nvarchar.ForVarcharStorage(); |
| 24 | + |
| 25 | + private static void CollectHashMismatch(Collation collation, string x, string y, List<string> mismatches) |
| 26 | + { |
| 27 | + if (collation.Equals(x, y) && collation.GetHashCode(x) != collation.GetHashCode(y)) |
| 28 | + mismatches.Add($"{Escape(x)} vs {Escape(y)} ({(ReferenceEquals(collation, Varchar) ? "varchar" : "nvarchar")})"); |
| 29 | + } |
| 30 | + |
| 31 | + private static string Escape(string s) => |
| 32 | + string.Concat(s.Select(c => c is >= ' ' and <= '~' ? c.ToString() : $"\\u{(int)c:X4}")); |
| 33 | + |
| 34 | + private static void AssertEqualAndHashEqual(Collation collation, string x, string y) |
| 35 | + { |
| 36 | + IsTrue(collation.Equals(x, y), $"expected Equals: '{x}' vs '{y}'"); |
| 37 | + AreEqual(collation.GetHashCode(x), collation.GetHashCode(y), $"hash differs: '{x}' vs '{y}'"); |
| 38 | + } |
| 39 | + |
| 40 | + [TestMethod] |
| 41 | + public void FullwidthSpelling_HashesLikeAsciiTarget() |
| 42 | + { |
| 43 | + foreach (var collation in (Collation[])[Nvarchar, Varchar]) |
| 44 | + { |
| 45 | + AssertEqualAndHashEqual(collation, "sp_executesql", "ｓp_executesql"); |
| 46 | + AssertEqualAndHashEqual(collation, "table1", "ｔable１"); |
| 47 | + AssertEqualAndHashEqual(collation, "ABC", "ａｂｃ"); |
| 48 | + } |
| 49 | + } |
| 50 | + |
| 51 | + [TestMethod] |
| 52 | + public void DecomposedAccent_HashesLikeComposed() |
| 53 | + { |
| 54 | + foreach (var collation in (Collation[])[Nvarchar, Varchar]) |
| 55 | + AssertEqualAndHashEqual(collation, "café", "cafe\u0301"); |
| 56 | + } |
| 57 | + |
| 58 | + [TestMethod] |
| 59 | + public void CapitalSharpS_HashesLikeSharpS() |
| 60 | + { |
| 61 | + foreach (var collation in (Collation[])[Nvarchar, Varchar]) |
| 62 | + AssertEqualAndHashEqual(collation, "straße", "straẞe"); |
| 63 | + } |
| 64 | + |
| 65 | + [TestMethod] |
| 66 | + public void IcuIgnorableCharacters_HashLikeAbsent() |
| 67 | + { |
| 68 | + foreach (var collation in (Collation[])[Nvarchar, Varchar]) |
| 69 | + { |
| 70 | + AssertEqualAndHashEqual(collation, "ab", "a\u200Bb"); |
| 71 | + AssertEqualAndHashEqual(collation, "ab", "a\uFEFFb"); |
| 72 | + // A control character is in-repertoire, so a pure-CP1252 pair |
| 73 | + // stays weight-compared (unequal); ICU ignores it only where |
| 74 | + // equality routes through the inner collation. The all-CP1252 |
| 75 | + // spelling still shares the hash — a legal collision the |
| 76 | + // fullwidth triangle requires. |
| 77 | + AssertEqualAndHashEqual(collation, "a\u0001ｂ", "aｂ"); |
| 78 | + AreEqual(collation.GetHashCode("a\u0001b"), collation.GetHashCode("ab")); |
| 79 | + } |
| 80 | + } |
| 81 | + |
| 82 | + /// <summary> |
| 83 | + /// An out-of-repertoire spelling can be Equals-equal to two |
| 84 | + /// in-repertoire strings that are unequal to each other (fullwidth |
| 85 | + /// <c>２</c> equals both <c>2</c> and superscript <c>²</c> through the |
| 86 | + /// inner collation), so all three must share a hash — the |
| 87 | + /// in-repertoire pair's shared hash is a legal collision of unequal |
| 88 | + /// strings. |
| 89 | + /// </summary> |
| 90 | + [TestMethod] |
| 91 | + public void CrossBoundaryTriangles_ShareOneHash() |
| 92 | + { |
| 93 | + AssertEqualAndHashEqual(Nvarchar, "x2y", "x２y"); |
| 94 | + AssertEqualAndHashEqual(Nvarchar, "x²y", "x２y"); |
| 95 | + AreEqual(Nvarchar.GetHashCode("x2y"), Nvarchar.GetHashCode("x²y")); |
| 96 | + |
| 97 | + AssertEqualAndHashEqual(Nvarchar, "a b", "a\u3000b"); |
| 98 | + AssertEqualAndHashEqual(Nvarchar, "a\u00A0b", "a\u3000b"); |
| 99 | + AreEqual(Nvarchar.GetHashCode("a b"), Nvarchar.GetHashCode("a\u00A0b")); |
| 100 | + |
| 101 | + // Thai SARA AM: ICU equates the composed U+0E33 with NIKHAHIT + |
| 102 | + // SARA AA while the weight tables keep the spellings unequal. |
| 103 | + AssertEqualAndHashEqual(Nvarchar, "กําx", "กำｘ"); |
| 104 | + AssertEqualAndHashEqual(Nvarchar, "กำx", "กำｘ"); |
| 105 | + AreEqual(Nvarchar.GetHashCode("กำx"), Nvarchar.GetHashCode("กําx")); |
| 106 | + } |
| 107 | + |
| 108 | + /// <summary> |
| 109 | + /// Completeness sweep: groups every repertoire character by its inner |
| 110 | + /// (invariant, CI + width/kana-insensitive) sort key and asserts each |
| 111 | + /// group's members hash identically in context — the behavioral form |
| 112 | + /// of "the in-repertoire hash-fold table covers every ICU-equal |
| 113 | + /// class" (case mates, superscripts, ordinal indicators, NBSP, Thai |
| 114 | + /// digits, ignorable controls). |
| 115 | + /// </summary> |
| 116 | + [TestMethod] |
| 117 | + public void InRepertoireIcuEqualClasses_ShareHashes() |
| 118 | + { |
| 119 | + const CompareOptions Options = |
| 120 | + CompareOptions.IgnoreCase | CompareOptions.IgnoreKanaType | CompareOptions.IgnoreWidth; |
| 121 | + var compareInfo = CultureInfo.InvariantCulture.CompareInfo; |
| 122 | + |
| 123 | + var groups = new Dictionary<string, List<char>>(StringComparer.Ordinal); |
| 124 | + foreach (var ch in Repertoire()) |
| 125 | + { |
| 126 | + var key = Convert.ToBase64String(compareInfo.GetSortKey(ch.ToString(), Options).KeyData); |
| 127 | + if (!groups.TryGetValue(key, out var members)) |
| 128 | + groups[key] = members = []; |
| 129 | + members.Add(ch); |
| 130 | + } |
| 131 | + |
| 132 | + var mismatches = new List<string>(); |
| 133 | + foreach (var members in groups.Values) |
| 134 | + { |
| 135 | + for (var i = 1; i < members.Count; i++) |
| 136 | + { |
| 137 | + foreach (var collation in (Collation[])[Nvarchar, Varchar]) |
| 138 | + { |
| 139 | + if (collation.GetHashCode($"x{members[0]}z") != collation.GetHashCode($"x{members[i]}z")) |
| 140 | + mismatches.Add($"U+{(int)members[0]:X4} vs U+{(int)members[i]:X4} ({(ReferenceEquals(collation, Varchar) ? "varchar" : "nvarchar")})"); |
| 141 | + } |
| 142 | + } |
| 143 | + } |
| 144 | + |
| 145 | + IsEmpty(mismatches, string.Join("; ", mismatches)); |
| 146 | + } |
| 147 | + |
| 148 | + /// <summary> |
| 149 | + /// Seeded fuzz over ICU-equivalence substitution pools: every generated |
| 150 | + /// pair that Equals reports equal must hash equal, under both storage |
| 151 | + /// bodies. Pools mix case flips, fullwidth forms, ignorables, |
| 152 | + /// decomposition, ligature case-mates, digit homoglyphs, and Thai. |
| 153 | + /// </summary> |
| 154 | + [TestMethod] |
| 155 | + public void SubstitutionFuzz_EqualsImpliesHashEqual() |
| 156 | + { |
| 157 | + string[][] pools = |
| 158 | + [ |
| 159 | + ["s", "S", "ｓ", "Ｓ"], |
| 160 | + ["2", "²", "๒", "２"], |
| 161 | + ["ß", "ẞ", "ss"], |
| 162 | + ["æ", "Æ", "ae"], |
| 163 | + ["é", "e\u0301", "É"], |
| 164 | + ["", "\u200B", "\u0001", "\u00AD"], |
| 165 | + [" ", "\u00A0", "\u3000"], |
| 166 | + ["ำ", "ํา"], |
| 167 | + ["-", "'", "_", "9", "ก", "z"], |
| 168 | + ["µ", "μ"], |
| 169 | + ]; |
| 170 | + |
| 171 | + var mismatches = new List<string>(); |
| 172 | + var random = new Random(20260713); |
| 173 | + for (var iteration = 0; iteration < 500; iteration++) |
| 174 | + { |
| 175 | + var length = random.Next(0, 6); |
| 176 | + var builderX = new StringBuilder(); |
| 177 | + var builderY = new StringBuilder(); |
| 178 | + for (var i = 0; i < length; i++) |
| 179 | + { |
| 180 | + var pool = pools[random.Next(pools.Length)]; |
| 181 | + _ = builderX.Append(pool[random.Next(pool.Length)]); |
| 182 | + _ = builderY.Append(pool[random.Next(pool.Length)]); |
| 183 | + } |
| 184 | + |
| 185 | + var x = builderX.ToString(); |
| 186 | + var y = builderY.ToString(); |
| 187 | + CollectHashMismatch(Nvarchar, x, y, mismatches); |
| 188 | + CollectHashMismatch(Varchar, x, y, mismatches); |
| 189 | + } |
| 190 | + |
| 191 | + IsEmpty(mismatches, string.Join("; ", mismatches)); |
| 192 | + } |
| 193 | + |
| 194 | + /// <summary> |
| 195 | + /// Systematic single-character sweep: for each BMP character in the |
| 196 | + /// blocks the fold machinery covers, derive normalization / case |
| 197 | + /// variants of a carrier string and assert the hash contract wherever |
| 198 | + /// Equals holds. |
| 199 | + /// </summary> |
| 200 | + [TestMethod] |
| 201 | + public void NormalizationVariantSweep_EqualsImpliesHashEqual() |
| 202 | + { |
| 203 | + (int Start, int End)[] blocks = |
| 204 | + [ |
| 205 | + (0x0020, 0x02FF), (0x0E00, 0x0E7F), (0x1E00, 0x1EFF), |
| 206 | + (0x2000, 0x20AF), (0x2100, 0x214F), (0xFB00, 0xFB06), (0xFF00, 0xFFEF), |
| 207 | + ]; |
| 208 | + var mismatches = new List<string>(); |
| 209 | + foreach (var (start, end) in blocks) |
| 210 | + { |
| 211 | + for (var cp = start; cp <= end; cp++) |
| 212 | + { |
| 213 | + var carrier = $"x{(char)cp}z"; |
| 214 | + foreach (var variant in Variants(carrier)) |
| 215 | + { |
| 216 | + if (variant == carrier) |
| 217 | + continue; |
| 218 | + CollectHashMismatch(Nvarchar, carrier, variant, mismatches); |
| 219 | + CollectHashMismatch(Varchar, carrier, variant, mismatches); |
| 220 | + } |
| 221 | + } |
| 222 | + } |
| 223 | + |
| 224 | + IsEmpty(mismatches, string.Join("; ", mismatches)); |
| 225 | + } |
| 226 | + |
| 227 | + private static IEnumerable<string> Variants(string s) |
| 228 | + { |
| 229 | + yield return s.ToUpperInvariant(); |
| 230 | + string[] normalized; |
| 231 | + try |
| 232 | + { |
| 233 | + normalized = |
| 234 | + [ |
| 235 | + s.Normalize(NormalizationForm.FormC), |
| 236 | + s.Normalize(NormalizationForm.FormD), |
| 237 | + s.Normalize(NormalizationForm.FormKC), |
| 238 | + s.Normalize(NormalizationForm.FormKD), |
| 239 | + ]; |
| 240 | + } |
| 241 | + catch (ArgumentException) |
| 242 | + { |
| 243 | + yield break; |
| 244 | + } |
| 245 | + |
| 246 | + foreach (var form in normalized) |
| 247 | + yield return form; |
| 248 | + } |
| 249 | + |
| 250 | + private static IEnumerable<char> Repertoire() |
| 251 | + { |
| 252 | + var encoding = CharSqlType.Cp1252Encoder; |
| 253 | + var buffer = new byte[1]; |
| 254 | + for (var b = 1; b <= 255; b++) |
| 255 | + { |
| 256 | + buffer[0] = (byte)b; |
| 257 | + var decoded = encoding.GetString(buffer); |
| 258 | + if (decoded.Length == 1) |
| 259 | + yield return decoded[0]; |
| 260 | + } |
| 261 | + |
| 262 | + for (var cp = 0x0E00; cp <= 0x0E7F; cp++) |
| 263 | + yield return (char)cp; |
| 264 | + } |
| 265 | +} |
0 commit comments