1 //===-- JSONTest.cpp - JSON unit tests --------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 #include "llvm/Support/JSON.h" 11 12 #include "gmock/gmock.h" 13 #include "gtest/gtest.h" 14 15 namespace llvm { 16 namespace json { 17 18 namespace { 19 20 std::string s(const Value &E) { return llvm::formatv("{0}", E).str(); } 21 std::string sp(const Value &E) { return llvm::formatv("{0:2}", E).str(); } 22 23 TEST(JSONTest, Types) { 24 EXPECT_EQ("true", s(true)); 25 EXPECT_EQ("null", s(nullptr)); 26 EXPECT_EQ("2.5", s(2.5)); 27 EXPECT_EQ(R"("foo")", s("foo")); 28 EXPECT_EQ("[1,2,3]", s({1, 2, 3})); 29 EXPECT_EQ(R"({"x":10,"y":20})", s(Object{{"x", 10}, {"y", 20}})); 30 31 #ifdef NDEBUG 32 EXPECT_EQ(R"("��")", s("\xC0\x80")); 33 EXPECT_EQ(R"({"��":0})", s(Object{{"\xC0\x80", 0}})); 34 #else 35 EXPECT_DEATH(s("\xC0\x80"), "Invalid UTF-8"); 36 EXPECT_DEATH(s(Object{{"\xC0\x80", 0}}), "Invalid UTF-8"); 37 #endif 38 } 39 40 TEST(JSONTest, Constructors) { 41 // Lots of edge cases around empty and singleton init lists. 42 EXPECT_EQ("[[[3]]]", s({{{3}}})); 43 EXPECT_EQ("[[[]]]", s({{{}}})); 44 EXPECT_EQ("[[{}]]", s({{Object{}}})); 45 EXPECT_EQ(R"({"A":{"B":{}}})", s(Object{{"A", Object{{"B", Object{}}}}})); 46 EXPECT_EQ(R"({"A":{"B":{"X":"Y"}}})", 47 s(Object{{"A", Object{{"B", Object{{"X", "Y"}}}}}})); 48 EXPECT_EQ("null", s(llvm::Optional<double>())); 49 EXPECT_EQ("2.5", s(llvm::Optional<double>(2.5))); 50 } 51 52 TEST(JSONTest, StringOwnership) { 53 char X[] = "Hello"; 54 Value Alias = static_cast<const char *>(X); 55 X[1] = 'a'; 56 EXPECT_EQ(R"("Hallo")", s(Alias)); 57 58 std::string Y = "Hello"; 59 Value Copy = Y; 60 Y[1] = 'a'; 61 EXPECT_EQ(R"("Hello")", s(Copy)); 62 } 63 64 TEST(JSONTest, CanonicalOutput) { 65 // Objects are sorted (but arrays aren't)! 66 EXPECT_EQ(R"({"a":1,"b":2,"c":3})", s(Object{{"a", 1}, {"c", 3}, {"b", 2}})); 67 EXPECT_EQ(R"(["a","c","b"])", s({"a", "c", "b"})); 68 EXPECT_EQ("3", s(3.0)); 69 } 70 71 TEST(JSONTest, Escaping) { 72 std::string test = { 73 0, // Strings may contain nulls. 74 '\b', '\f', // Have mnemonics, but we escape numerically. 75 '\r', '\n', '\t', // Escaped with mnemonics. 76 'S', '\"', '\\', // Printable ASCII characters. 77 '\x7f', // Delete is not escaped. 78 '\xce', '\x94', // Non-ASCII UTF-8 is not escaped. 79 }; 80 81 std::string teststring = R"("\u0000\u0008\u000c\r\n\tS\"\\)" 82 "\x7f\xCE\x94\""; 83 84 EXPECT_EQ(teststring, s(test)); 85 86 EXPECT_EQ(R"({"object keys are\nescaped":true})", 87 s(Object{{"object keys are\nescaped", true}})); 88 } 89 90 TEST(JSONTest, PrettyPrinting) { 91 const char str[] = R"({ 92 "empty_array": [], 93 "empty_object": {}, 94 "full_array": [ 95 1, 96 null 97 ], 98 "full_object": { 99 "nested_array": [ 100 { 101 "property": "value" 102 } 103 ] 104 } 105 })"; 106 107 EXPECT_EQ(str, sp(Object{ 108 {"empty_object", Object{}}, 109 {"empty_array", {}}, 110 {"full_array", {1, nullptr}}, 111 {"full_object", 112 Object{ 113 {"nested_array", 114 {Object{ 115 {"property", "value"}, 116 }}}, 117 }}, 118 })); 119 } 120 121 TEST(JSONTest, Parse) { 122 auto Compare = [](llvm::StringRef S, Value Expected) { 123 if (auto E = parse(S)) { 124 // Compare both string forms and with operator==, in case we have bugs. 125 EXPECT_EQ(*E, Expected); 126 EXPECT_EQ(sp(*E), sp(Expected)); 127 } else { 128 handleAllErrors(E.takeError(), [S](const llvm::ErrorInfoBase &E) { 129 FAIL() << "Failed to parse JSON >>> " << S << " <<<: " << E.message(); 130 }); 131 } 132 }; 133 134 Compare(R"(true)", true); 135 Compare(R"(false)", false); 136 Compare(R"(null)", nullptr); 137 138 Compare(R"(42)", 42); 139 Compare(R"(2.5)", 2.5); 140 Compare(R"(2e50)", 2e50); 141 Compare(R"(1.2e3456789)", std::numeric_limits<double>::infinity()); 142 143 Compare(R"("foo")", "foo"); 144 Compare(R"("\"\\\b\f\n\r\t")", "\"\\\b\f\n\r\t"); 145 Compare(R"("\u0000")", llvm::StringRef("\0", 1)); 146 Compare("\"\x7f\"", "\x7f"); 147 Compare(R"("\ud801\udc37")", u8"\U00010437"); // UTF16 surrogate pair escape. 148 Compare("\"\xE2\x82\xAC\xF0\x9D\x84\x9E\"", u8"\u20ac\U0001d11e"); // UTF8 149 Compare( 150 R"("LoneLeading=\ud801, LoneTrailing=\udc01, LeadingLeadingTrailing=\ud801\ud801\udc37")", 151 u8"LoneLeading=\ufffd, LoneTrailing=\ufffd, " 152 u8"LeadingLeadingTrailing=\ufffd\U00010437"); // Invalid unicode. 153 154 Compare(R"({"":0,"":0})", Object{{"", 0}}); 155 Compare(R"({"obj":{},"arr":[]})", Object{{"obj", Object{}}, {"arr", {}}}); 156 Compare(R"({"\n":{"\u0000":[[[[]]]]}})", 157 Object{{"\n", Object{ 158 {llvm::StringRef("\0", 1), {{{{}}}}}, 159 }}}); 160 Compare("\r[\n\t] ", {}); 161 } 162 163 TEST(JSONTest, ParseErrors) { 164 auto ExpectErr = [](llvm::StringRef Msg, llvm::StringRef S) { 165 if (auto E = parse(S)) { 166 // Compare both string forms and with operator==, in case we have bugs. 167 FAIL() << "Parsed JSON >>> " << S << " <<< but wanted error: " << Msg; 168 } else { 169 handleAllErrors(E.takeError(), [S, Msg](const llvm::ErrorInfoBase &E) { 170 EXPECT_THAT(E.message(), testing::HasSubstr(Msg)) << S; 171 }); 172 } 173 }; 174 ExpectErr("Unexpected EOF", ""); 175 ExpectErr("Unexpected EOF", "["); 176 ExpectErr("Text after end of document", "[][]"); 177 ExpectErr("Invalid JSON value (false?)", "fuzzy"); 178 ExpectErr("Expected , or ]", "[2?]"); 179 ExpectErr("Expected object key", "{a:2}"); 180 ExpectErr("Expected : after object key", R"({"a",2})"); 181 ExpectErr("Expected , or } after object property", R"({"a":2 "b":3})"); 182 ExpectErr("Invalid JSON value", R"([&%!])"); 183 ExpectErr("Invalid JSON value (number?)", "1e1.0"); 184 ExpectErr("Unterminated string", R"("abc\"def)"); 185 ExpectErr("Control character in string", "\"abc\ndef\""); 186 ExpectErr("Invalid escape sequence", R"("\030")"); 187 ExpectErr("Invalid \\u escape sequence", R"("\usuck")"); 188 ExpectErr("[3:3, byte=19]", R"({ 189 "valid": 1, 190 invalid: 2 191 })"); 192 ExpectErr("Invalid UTF-8 sequence", "\"\xC0\x80\""); // WTF-8 null 193 } 194 195 // Direct tests of isUTF8 and fixUTF8. Internal uses are also tested elsewhere. 196 TEST(JSONTest, UTF8) { 197 for (const char *Valid : { 198 "this is ASCII text", 199 "thïs tëxt häs BMP chäräctërs", 200 "����L���� C��������", 201 }) { 202 EXPECT_TRUE(isUTF8(Valid)) << Valid; 203 EXPECT_EQ(fixUTF8(Valid), Valid); 204 } 205 for (auto Invalid : std::vector<std::pair<const char *, const char *>>{ 206 {"lone trailing \x81\x82 bytes", "lone trailing �� bytes"}, 207 {"missing trailing \xD0 bytes", "missing trailing � bytes"}, 208 {"truncated character \xD0", "truncated character �"}, 209 {"not \xC1\x80 the \xE0\x9f\xBF shortest \xF0\x83\x83\x83 encoding", 210 "not �� the ��� shortest ���� encoding"}, 211 {"too \xF9\x80\x80\x80\x80 long", "too ����� long"}, 212 {"surrogate \xED\xA0\x80 invalid \xF4\x90\x80\x80", 213 "surrogate ��� invalid ����"}}) { 214 EXPECT_FALSE(isUTF8(Invalid.first)) << Invalid.first; 215 EXPECT_EQ(fixUTF8(Invalid.first), Invalid.second); 216 } 217 } 218 219 TEST(JSONTest, Inspection) { 220 llvm::Expected<Value> Doc = parse(R"( 221 { 222 "null": null, 223 "boolean": false, 224 "number": 2.78, 225 "string": "json", 226 "array": [null, true, 3.14, "hello", [1,2,3], {"time": "arrow"}], 227 "object": {"fruit": "banana"} 228 } 229 )"); 230 EXPECT_TRUE(!!Doc); 231 232 Object *O = Doc->getAsObject(); 233 ASSERT_TRUE(O); 234 235 EXPECT_FALSE(O->getNull("missing")); 236 EXPECT_FALSE(O->getNull("boolean")); 237 EXPECT_TRUE(O->getNull("null")); 238 239 EXPECT_EQ(O->getNumber("number"), llvm::Optional<double>(2.78)); 240 EXPECT_FALSE(O->getInteger("number")); 241 EXPECT_EQ(O->getString("string"), llvm::Optional<llvm::StringRef>("json")); 242 ASSERT_FALSE(O->getObject("missing")); 243 ASSERT_FALSE(O->getObject("array")); 244 ASSERT_TRUE(O->getObject("object")); 245 EXPECT_EQ(*O->getObject("object"), (Object{{"fruit", "banana"}})); 246 247 Array *A = O->getArray("array"); 248 ASSERT_TRUE(A); 249 EXPECT_EQ((*A)[1].getAsBoolean(), llvm::Optional<bool>(true)); 250 ASSERT_TRUE((*A)[4].getAsArray()); 251 EXPECT_EQ(*(*A)[4].getAsArray(), (Array{1, 2, 3})); 252 EXPECT_EQ((*(*A)[4].getAsArray())[1].getAsInteger(), 253 llvm::Optional<int64_t>(2)); 254 int I = 0; 255 for (Value &E : *A) { 256 if (I++ == 5) { 257 ASSERT_TRUE(E.getAsObject()); 258 EXPECT_EQ(E.getAsObject()->getString("time"), 259 llvm::Optional<llvm::StringRef>("arrow")); 260 } else 261 EXPECT_FALSE(E.getAsObject()); 262 } 263 } 264 265 // Verify special integer handling - we try to preserve exact int64 values. 266 TEST(JSONTest, Integers) { 267 struct { 268 const char *Desc; 269 Value Val; 270 const char *Str; 271 llvm::Optional<int64_t> AsInt; 272 llvm::Optional<double> AsNumber; 273 } TestCases[] = { 274 { 275 "Non-integer. Stored as double, not convertible.", 276 double{1.5}, 277 "1.5", 278 llvm::None, 279 1.5, 280 }, 281 282 { 283 "Integer, not exact double. Stored as int64, convertible.", 284 int64_t{0x4000000000000001}, 285 "4611686018427387905", 286 int64_t{0x4000000000000001}, 287 double{0x4000000000000000}, 288 }, 289 290 { 291 "Negative integer, not exact double. Stored as int64, convertible.", 292 int64_t{-0x4000000000000001}, 293 "-4611686018427387905", 294 int64_t{-0x4000000000000001}, 295 double{-0x4000000000000000}, 296 }, 297 298 { 299 "Dynamically exact integer. Stored as double, convertible.", 300 double{0x6000000000000000}, 301 "6.9175290276410819e+18", 302 int64_t{0x6000000000000000}, 303 double{0x6000000000000000}, 304 }, 305 306 { 307 "Dynamically integer, >64 bits. Stored as double, not convertible.", 308 1.5 * double{0x8000000000000000}, 309 "1.3835058055282164e+19", 310 llvm::None, 311 1.5 * double{0x8000000000000000}, 312 }, 313 }; 314 for (const auto &T : TestCases) { 315 EXPECT_EQ(T.Str, s(T.Val)) << T.Desc; 316 llvm::Expected<Value> Doc = parse(T.Str); 317 EXPECT_TRUE(!!Doc) << T.Desc; 318 EXPECT_EQ(Doc->getAsInteger(), T.AsInt) << T.Desc; 319 EXPECT_EQ(Doc->getAsNumber(), T.AsNumber) << T.Desc; 320 EXPECT_EQ(T.Val, *Doc) << T.Desc; 321 EXPECT_EQ(T.Str, s(*Doc)) << T.Desc; 322 } 323 } 324 325 // Sample struct with typical JSON-mapping rules. 326 struct CustomStruct { 327 CustomStruct() : B(false) {} 328 CustomStruct(std::string S, llvm::Optional<int> I, bool B) 329 : S(S), I(I), B(B) {} 330 std::string S; 331 llvm::Optional<int> I; 332 bool B; 333 }; 334 inline bool operator==(const CustomStruct &L, const CustomStruct &R) { 335 return L.S == R.S && L.I == R.I && L.B == R.B; 336 } 337 inline llvm::raw_ostream &operator<<(llvm::raw_ostream &OS, 338 const CustomStruct &S) { 339 return OS << "(" << S.S << ", " << (S.I ? std::to_string(*S.I) : "None") 340 << ", " << S.B << ")"; 341 } 342 bool fromJSON(const Value &E, CustomStruct &R) { 343 ObjectMapper O(E); 344 if (!O || !O.map("str", R.S) || !O.map("int", R.I)) 345 return false; 346 O.map("bool", R.B); 347 return true; 348 } 349 350 TEST(JSONTest, Deserialize) { 351 std::map<std::string, std::vector<CustomStruct>> R; 352 CustomStruct ExpectedStruct = {"foo", 42, true}; 353 std::map<std::string, std::vector<CustomStruct>> Expected; 354 Value J = Object{ 355 {"foo", 356 Array{ 357 Object{ 358 {"str", "foo"}, 359 {"int", 42}, 360 {"bool", true}, 361 {"unknown", "ignored"}, 362 }, 363 Object{{"str", "bar"}}, 364 Object{ 365 {"str", "baz"}, {"bool", "string"}, // OK, deserialize ignores. 366 }, 367 }}}; 368 Expected["foo"] = { 369 CustomStruct("foo", 42, true), 370 CustomStruct("bar", llvm::None, false), 371 CustomStruct("baz", llvm::None, false), 372 }; 373 ASSERT_TRUE(fromJSON(J, R)); 374 EXPECT_EQ(R, Expected); 375 376 CustomStruct V; 377 EXPECT_FALSE(fromJSON(nullptr, V)) << "Not an object " << V; 378 EXPECT_FALSE(fromJSON(Object{}, V)) << "Missing required field " << V; 379 EXPECT_FALSE(fromJSON(Object{{"str", 1}}, V)) << "Wrong type " << V; 380 // Optional<T> must parse as the correct type if present. 381 EXPECT_FALSE(fromJSON(Object{{"str", 1}, {"int", "string"}}, V)) 382 << "Wrong type for Optional<T> " << V; 383 } 384 385 } // namespace 386 } // namespace json 387 } // namespace llvm 388