| 1 | #include <Core/Defines.h> |
| 2 | #include <Common/hex.h> |
| 3 | #include <Common/PODArray.h> |
| 4 | #include <Common/StringUtils/StringUtils.h> |
| 5 | #include <Common/memcpySmall.h> |
| 6 | #include <Formats/FormatSettings.h> |
| 7 | #include <IO/WriteHelpers.h> |
| 8 | #include <IO/WriteBufferFromString.h> |
| 9 | #include <IO/readFloatText.h> |
| 10 | #include <IO/Operators.h> |
| 11 | #include <common/find_symbols.h> |
| 12 | #include <stdlib.h> |
| 13 | |
| 14 | #ifdef __SSE2__ |
| 15 | #include <emmintrin.h> |
| 16 | #endif |
| 17 | |
| 18 | namespace DB |
| 19 | { |
| 20 | |
| 21 | namespace ErrorCodes |
| 22 | { |
| 23 | extern const int CANNOT_PARSE_INPUT_ASSERTION_FAILED; |
| 24 | extern const int CANNOT_PARSE_ESCAPE_SEQUENCE; |
| 25 | extern const int CANNOT_PARSE_QUOTED_STRING; |
| 26 | extern const int CANNOT_PARSE_DATETIME; |
| 27 | extern const int CANNOT_PARSE_DATE; |
| 28 | extern const int INCORRECT_DATA; |
| 29 | } |
| 30 | |
| 31 | template <typename IteratorSrc, typename IteratorDst> |
| 32 | void parseHex(IteratorSrc src, IteratorDst dst, const size_t num_bytes) |
| 33 | { |
| 34 | size_t src_pos = 0; |
| 35 | size_t dst_pos = 0; |
| 36 | for (; dst_pos < num_bytes; ++dst_pos) |
| 37 | { |
| 38 | dst[dst_pos] = UInt8(unhex(src[src_pos])) * 16 + UInt8(unhex(src[src_pos + 1])); |
| 39 | src_pos += 2; |
| 40 | } |
| 41 | } |
| 42 | |
| 43 | void parseUUID(const UInt8 * src36, UInt8 * dst16) |
| 44 | { |
| 45 | /// If string is not like UUID - implementation specific behaviour. |
| 46 | |
| 47 | parseHex(&src36[0], &dst16[0], 4); |
| 48 | parseHex(&src36[9], &dst16[4], 2); |
| 49 | parseHex(&src36[14], &dst16[6], 2); |
| 50 | parseHex(&src36[19], &dst16[8], 2); |
| 51 | parseHex(&src36[24], &dst16[10], 6); |
| 52 | } |
| 53 | |
| 54 | /** Function used when byte ordering is important when parsing uuid |
| 55 | * ex: When we create an UUID type |
| 56 | */ |
| 57 | void parseUUID(const UInt8 * src36, std::reverse_iterator<UInt8 *> dst16) |
| 58 | { |
| 59 | /// If string is not like UUID - implementation specific behaviour. |
| 60 | |
| 61 | /// FIXME This code looks like trash. |
| 62 | parseHex(&src36[0], dst16 + 8, 4); |
| 63 | parseHex(&src36[9], dst16 + 12, 2); |
| 64 | parseHex(&src36[14], dst16 + 14, 2); |
| 65 | parseHex(&src36[19], dst16, 2); |
| 66 | parseHex(&src36[24], dst16 + 2, 6); |
| 67 | } |
| 68 | |
| 69 | UInt128 stringToUUID(const String & str) |
| 70 | { |
| 71 | return parseFromString<UUID>(str); |
| 72 | } |
| 73 | |
| 74 | void NO_INLINE throwAtAssertionFailed(const char * s, ReadBuffer & buf) |
| 75 | { |
| 76 | WriteBufferFromOwnString out; |
| 77 | out << "Cannot parse input: expected " << escape << s; |
| 78 | |
| 79 | if (buf.eof()) |
| 80 | out << " at end of stream." ; |
| 81 | else |
| 82 | out << " before: " << escape << String(buf.position(), std::min(SHOW_CHARS_ON_SYNTAX_ERROR, buf.buffer().end() - buf.position())); |
| 83 | |
| 84 | throw Exception(out.str(), ErrorCodes::CANNOT_PARSE_INPUT_ASSERTION_FAILED); |
| 85 | } |
| 86 | |
| 87 | |
| 88 | bool checkString(const char * s, ReadBuffer & buf) |
| 89 | { |
| 90 | for (; *s; ++s) |
| 91 | { |
| 92 | if (buf.eof() || *buf.position() != *s) |
| 93 | return false; |
| 94 | ++buf.position(); |
| 95 | } |
| 96 | return true; |
| 97 | } |
| 98 | |
| 99 | |
| 100 | bool checkStringCaseInsensitive(const char * s, ReadBuffer & buf) |
| 101 | { |
| 102 | for (; *s; ++s) |
| 103 | { |
| 104 | if (buf.eof()) |
| 105 | return false; |
| 106 | |
| 107 | char c = *buf.position(); |
| 108 | if (!equalsCaseInsensitive(*s, c)) |
| 109 | return false; |
| 110 | |
| 111 | ++buf.position(); |
| 112 | } |
| 113 | return true; |
| 114 | } |
| 115 | |
| 116 | |
| 117 | void assertString(const char * s, ReadBuffer & buf) |
| 118 | { |
| 119 | if (!checkString(s, buf)) |
| 120 | throwAtAssertionFailed(s, buf); |
| 121 | } |
| 122 | |
| 123 | |
| 124 | void assertEOF(ReadBuffer & buf) |
| 125 | { |
| 126 | if (!buf.eof()) |
| 127 | throwAtAssertionFailed("eof" , buf); |
| 128 | } |
| 129 | |
| 130 | |
| 131 | void assertStringCaseInsensitive(const char * s, ReadBuffer & buf) |
| 132 | { |
| 133 | if (!checkStringCaseInsensitive(s, buf)) |
| 134 | throwAtAssertionFailed(s, buf); |
| 135 | } |
| 136 | |
| 137 | |
| 138 | bool checkStringByFirstCharacterAndAssertTheRest(const char * s, ReadBuffer & buf) |
| 139 | { |
| 140 | if (buf.eof() || *buf.position() != *s) |
| 141 | return false; |
| 142 | |
| 143 | assertString(s, buf); |
| 144 | return true; |
| 145 | } |
| 146 | |
| 147 | bool checkStringByFirstCharacterAndAssertTheRestCaseInsensitive(const char * s, ReadBuffer & buf) |
| 148 | { |
| 149 | if (buf.eof()) |
| 150 | return false; |
| 151 | |
| 152 | char c = *buf.position(); |
| 153 | if (!equalsCaseInsensitive(*s, c)) |
| 154 | return false; |
| 155 | |
| 156 | assertStringCaseInsensitive(s, buf); |
| 157 | return true; |
| 158 | } |
| 159 | |
| 160 | |
| 161 | template <typename T> |
| 162 | static void appendToStringOrVector(T & s, ReadBuffer & rb, const char * end) |
| 163 | { |
| 164 | s.append(rb.position(), end - rb.position()); |
| 165 | } |
| 166 | |
| 167 | template <> |
| 168 | inline void appendToStringOrVector(PaddedPODArray<UInt8> & s, ReadBuffer & rb, const char * end) |
| 169 | { |
| 170 | if (rb.isPadded()) |
| 171 | s.insertSmallAllowReadWriteOverflow15(rb.position(), end); |
| 172 | else |
| 173 | s.insert(rb.position(), end); |
| 174 | } |
| 175 | |
| 176 | template <> |
| 177 | inline void appendToStringOrVector(PODArray<char> & s, ReadBuffer & rb, const char * end) |
| 178 | { |
| 179 | s.insert(rb.position(), end); |
| 180 | } |
| 181 | |
| 182 | template <typename Vector> |
| 183 | void readStringInto(Vector & s, ReadBuffer & buf) |
| 184 | { |
| 185 | while (!buf.eof()) |
| 186 | { |
| 187 | char * next_pos = find_first_symbols<'\t', '\n'>(buf.position(), buf.buffer().end()); |
| 188 | |
| 189 | appendToStringOrVector(s, buf, next_pos); |
| 190 | buf.position() = next_pos; |
| 191 | |
| 192 | if (buf.hasPendingData()) |
| 193 | return; |
| 194 | } |
| 195 | } |
| 196 | |
| 197 | template <typename Vector> |
| 198 | void readNullTerminated(Vector & s, ReadBuffer & buf) |
| 199 | { |
| 200 | while (!buf.eof()) |
| 201 | { |
| 202 | char * next_pos = find_first_symbols<'\0'>(buf.position(), buf.buffer().end()); |
| 203 | |
| 204 | appendToStringOrVector(s, buf, next_pos); |
| 205 | buf.position() = next_pos; |
| 206 | |
| 207 | if (buf.hasPendingData()) |
| 208 | break; |
| 209 | } |
| 210 | buf.ignore(); |
| 211 | } |
| 212 | |
| 213 | template void readNullTerminated<PODArray<char>>(PODArray<char> & s, ReadBuffer & buf); |
| 214 | template void readNullTerminated<String>(String & s, ReadBuffer & buf); |
| 215 | |
| 216 | void readString(String & s, ReadBuffer & buf) |
| 217 | { |
| 218 | s.clear(); |
| 219 | readStringInto(s, buf); |
| 220 | } |
| 221 | |
| 222 | template void readStringInto<PaddedPODArray<UInt8>>(PaddedPODArray<UInt8> & s, ReadBuffer & buf); |
| 223 | |
| 224 | |
| 225 | template <typename Vector> |
| 226 | void readStringUntilEOFInto(Vector & s, ReadBuffer & buf) |
| 227 | { |
| 228 | while (!buf.eof()) |
| 229 | { |
| 230 | appendToStringOrVector(s, buf, buf.buffer().end()); |
| 231 | buf.position() = buf.buffer().end(); |
| 232 | |
| 233 | if (buf.hasPendingData()) |
| 234 | return; |
| 235 | } |
| 236 | } |
| 237 | |
| 238 | |
| 239 | void readStringUntilEOF(String & s, ReadBuffer & buf) |
| 240 | { |
| 241 | s.clear(); |
| 242 | readStringUntilEOFInto(s, buf); |
| 243 | } |
| 244 | |
| 245 | template <typename Vector> |
| 246 | void readEscapedStringUntilEOLInto(Vector & s, ReadBuffer & buf) |
| 247 | { |
| 248 | while (!buf.eof()) |
| 249 | { |
| 250 | char * next_pos = find_first_symbols<'\n', '\\'>(buf.position(), buf.buffer().end()); |
| 251 | |
| 252 | appendToStringOrVector(s, buf, next_pos); |
| 253 | buf.position() = next_pos; |
| 254 | |
| 255 | if (!buf.hasPendingData()) |
| 256 | continue; |
| 257 | |
| 258 | if (*buf.position() == '\n') |
| 259 | return; |
| 260 | |
| 261 | if (*buf.position() == '\\') |
| 262 | parseComplexEscapeSequence(s, buf); |
| 263 | } |
| 264 | } |
| 265 | |
| 266 | |
| 267 | void readEscapedStringUntilEOL(String & s, ReadBuffer & buf) |
| 268 | { |
| 269 | s.clear(); |
| 270 | readEscapedStringUntilEOLInto(s, buf); |
| 271 | } |
| 272 | |
| 273 | template void readStringUntilEOFInto<PaddedPODArray<UInt8>>(PaddedPODArray<UInt8> & s, ReadBuffer & buf); |
| 274 | |
| 275 | |
| 276 | /** Parse the escape sequence, which can be simple (one character after backslash) or more complex (multiple characters). |
| 277 | * It is assumed that the cursor is located on the `\` symbol |
| 278 | */ |
| 279 | template <typename Vector> |
| 280 | static void parseComplexEscapeSequence(Vector & s, ReadBuffer & buf) |
| 281 | { |
| 282 | ++buf.position(); |
| 283 | if (buf.eof()) |
| 284 | throw Exception("Cannot parse escape sequence" , ErrorCodes::CANNOT_PARSE_ESCAPE_SEQUENCE); |
| 285 | |
| 286 | if (*buf.position() == 'x') |
| 287 | { |
| 288 | ++buf.position(); |
| 289 | /// escape sequence of the form \xAA |
| 290 | char hex_code[2]; |
| 291 | readPODBinary(hex_code, buf); |
| 292 | s.push_back(unhex2(hex_code)); |
| 293 | } |
| 294 | else if (*buf.position() == 'N') |
| 295 | { |
| 296 | /// Support for NULLs: \N sequence must be parsed as empty string. |
| 297 | ++buf.position(); |
| 298 | } |
| 299 | else |
| 300 | { |
| 301 | /// The usual escape sequence of a single character. |
| 302 | s.push_back(parseEscapeSequence(*buf.position())); |
| 303 | ++buf.position(); |
| 304 | } |
| 305 | } |
| 306 | |
| 307 | |
| 308 | template <typename Vector, typename ReturnType> |
| 309 | static ReturnType parseJSONEscapeSequence(Vector & s, ReadBuffer & buf) |
| 310 | { |
| 311 | static constexpr bool throw_exception = std::is_same_v<ReturnType, void>; |
| 312 | |
| 313 | auto error = [](const char * message [[maybe_unused]], int code [[maybe_unused]]) |
| 314 | { |
| 315 | if constexpr (throw_exception) |
| 316 | throw Exception(message, code); |
| 317 | return ReturnType(false); |
| 318 | }; |
| 319 | |
| 320 | ++buf.position(); |
| 321 | if (buf.eof()) |
| 322 | return error("Cannot parse escape sequence" , ErrorCodes::CANNOT_PARSE_ESCAPE_SEQUENCE); |
| 323 | |
| 324 | switch (*buf.position()) |
| 325 | { |
| 326 | case '"': |
| 327 | s.push_back('"'); |
| 328 | break; |
| 329 | case '\\': |
| 330 | s.push_back('\\'); |
| 331 | break; |
| 332 | case '/': |
| 333 | s.push_back('/'); |
| 334 | break; |
| 335 | case 'b': |
| 336 | s.push_back('\b'); |
| 337 | break; |
| 338 | case 'f': |
| 339 | s.push_back('\f'); |
| 340 | break; |
| 341 | case 'n': |
| 342 | s.push_back('\n'); |
| 343 | break; |
| 344 | case 'r': |
| 345 | s.push_back('\r'); |
| 346 | break; |
| 347 | case 't': |
| 348 | s.push_back('\t'); |
| 349 | break; |
| 350 | case 'u': |
| 351 | { |
| 352 | ++buf.position(); |
| 353 | |
| 354 | char hex_code[4]; |
| 355 | if (4 != buf.read(hex_code, 4)) |
| 356 | return error("Cannot parse escape sequence: less than four bytes after \\u" , ErrorCodes::CANNOT_PARSE_ESCAPE_SEQUENCE); |
| 357 | |
| 358 | /// \u0000 - special case |
| 359 | if (0 == memcmp(hex_code, "0000" , 4)) |
| 360 | { |
| 361 | s.push_back(0); |
| 362 | return ReturnType(true); |
| 363 | } |
| 364 | |
| 365 | UInt16 code_point = unhex4(hex_code); |
| 366 | |
| 367 | if (code_point <= 0x7F) |
| 368 | { |
| 369 | s.push_back(code_point); |
| 370 | } |
| 371 | else if (code_point <= 0x07FF) |
| 372 | { |
| 373 | s.push_back(((code_point >> 6) & 0x1F) | 0xC0); |
| 374 | s.push_back((code_point & 0x3F) | 0x80); |
| 375 | } |
| 376 | else |
| 377 | { |
| 378 | /// Surrogate pair. |
| 379 | if (code_point >= 0xD800 && code_point <= 0xDBFF) |
| 380 | { |
| 381 | if (!checkString("\\u" , buf)) |
| 382 | return error("Cannot parse escape sequence: missing second part of surrogate pair" , ErrorCodes::CANNOT_PARSE_ESCAPE_SEQUENCE); |
| 383 | |
| 384 | char second_hex_code[4]; |
| 385 | if (4 != buf.read(second_hex_code, 4)) |
| 386 | return error("Cannot parse escape sequence: less than four bytes after \\u of second part of surrogate pair" , |
| 387 | ErrorCodes::CANNOT_PARSE_ESCAPE_SEQUENCE); |
| 388 | |
| 389 | UInt16 second_code_point = unhex4(second_hex_code); |
| 390 | |
| 391 | if (second_code_point >= 0xDC00 && second_code_point <= 0xDFFF) |
| 392 | { |
| 393 | UInt32 full_code_point = 0x10000 + (code_point - 0xD800) * 1024 + (second_code_point - 0xDC00); |
| 394 | |
| 395 | s.push_back(((full_code_point >> 18) & 0x07) | 0xF0); |
| 396 | s.push_back(((full_code_point >> 12) & 0x3F) | 0x80); |
| 397 | s.push_back(((full_code_point >> 6) & 0x3F) | 0x80); |
| 398 | s.push_back((full_code_point & 0x3F) | 0x80); |
| 399 | } |
| 400 | else |
| 401 | return error("Incorrect surrogate pair of unicode escape sequences in JSON" , ErrorCodes::CANNOT_PARSE_ESCAPE_SEQUENCE); |
| 402 | } |
| 403 | else |
| 404 | { |
| 405 | s.push_back(((code_point >> 12) & 0x0F) | 0xE0); |
| 406 | s.push_back(((code_point >> 6) & 0x3F) | 0x80); |
| 407 | s.push_back((code_point & 0x3F) | 0x80); |
| 408 | } |
| 409 | } |
| 410 | |
| 411 | return ReturnType(true); |
| 412 | } |
| 413 | default: |
| 414 | s.push_back(*buf.position()); |
| 415 | break; |
| 416 | } |
| 417 | |
| 418 | ++buf.position(); |
| 419 | return ReturnType(true); |
| 420 | } |
| 421 | |
| 422 | |
| 423 | template <typename Vector> |
| 424 | void readEscapedStringInto(Vector & s, ReadBuffer & buf) |
| 425 | { |
| 426 | while (!buf.eof()) |
| 427 | { |
| 428 | char * next_pos = find_first_symbols<'\t', '\n', '\\'>(buf.position(), buf.buffer().end()); |
| 429 | |
| 430 | appendToStringOrVector(s, buf, next_pos); |
| 431 | buf.position() = next_pos; |
| 432 | |
| 433 | if (!buf.hasPendingData()) |
| 434 | continue; |
| 435 | |
| 436 | if (*buf.position() == '\t' || *buf.position() == '\n') |
| 437 | return; |
| 438 | |
| 439 | if (*buf.position() == '\\') |
| 440 | parseComplexEscapeSequence(s, buf); |
| 441 | } |
| 442 | } |
| 443 | |
| 444 | void readEscapedString(String & s, ReadBuffer & buf) |
| 445 | { |
| 446 | s.clear(); |
| 447 | readEscapedStringInto(s, buf); |
| 448 | } |
| 449 | |
| 450 | template void readEscapedStringInto<PaddedPODArray<UInt8>>(PaddedPODArray<UInt8> & s, ReadBuffer & buf); |
| 451 | template void readEscapedStringInto<NullSink>(NullSink & s, ReadBuffer & buf); |
| 452 | |
| 453 | |
| 454 | /** If enable_sql_style_quoting == true, |
| 455 | * strings like 'abc''def' will be parsed as abc'def. |
| 456 | * Please note, that even with SQL style quoting enabled, |
| 457 | * backslash escape sequences are also parsed, |
| 458 | * that could be slightly confusing. |
| 459 | */ |
| 460 | template <char quote, bool enable_sql_style_quoting, typename Vector> |
| 461 | static void readAnyQuotedStringInto(Vector & s, ReadBuffer & buf) |
| 462 | { |
| 463 | if (buf.eof() || *buf.position() != quote) |
| 464 | throw Exception("Cannot parse quoted string: expected opening quote" , |
| 465 | ErrorCodes::CANNOT_PARSE_QUOTED_STRING); |
| 466 | ++buf.position(); |
| 467 | |
| 468 | while (!buf.eof()) |
| 469 | { |
| 470 | char * next_pos = find_first_symbols<'\\', quote>(buf.position(), buf.buffer().end()); |
| 471 | |
| 472 | appendToStringOrVector(s, buf, next_pos); |
| 473 | buf.position() = next_pos; |
| 474 | |
| 475 | if (!buf.hasPendingData()) |
| 476 | continue; |
| 477 | |
| 478 | if (*buf.position() == quote) |
| 479 | { |
| 480 | ++buf.position(); |
| 481 | |
| 482 | if (enable_sql_style_quoting && !buf.eof() && *buf.position() == quote) |
| 483 | { |
| 484 | s.push_back(quote); |
| 485 | ++buf.position(); |
| 486 | continue; |
| 487 | } |
| 488 | |
| 489 | return; |
| 490 | } |
| 491 | |
| 492 | if (*buf.position() == '\\') |
| 493 | parseComplexEscapeSequence(s, buf); |
| 494 | } |
| 495 | |
| 496 | throw Exception("Cannot parse quoted string: expected closing quote" , |
| 497 | ErrorCodes::CANNOT_PARSE_QUOTED_STRING); |
| 498 | } |
| 499 | |
| 500 | template <bool enable_sql_style_quoting, typename Vector> |
| 501 | void readQuotedStringInto(Vector & s, ReadBuffer & buf) |
| 502 | { |
| 503 | readAnyQuotedStringInto<'\'', enable_sql_style_quoting>(s, buf); |
| 504 | } |
| 505 | |
| 506 | template <bool enable_sql_style_quoting, typename Vector> |
| 507 | void readDoubleQuotedStringInto(Vector & s, ReadBuffer & buf) |
| 508 | { |
| 509 | readAnyQuotedStringInto<'"', enable_sql_style_quoting>(s, buf); |
| 510 | } |
| 511 | |
| 512 | template <bool enable_sql_style_quoting, typename Vector> |
| 513 | void readBackQuotedStringInto(Vector & s, ReadBuffer & buf) |
| 514 | { |
| 515 | readAnyQuotedStringInto<'`', enable_sql_style_quoting>(s, buf); |
| 516 | } |
| 517 | |
| 518 | |
| 519 | void readQuotedString(String & s, ReadBuffer & buf) |
| 520 | { |
| 521 | s.clear(); |
| 522 | readQuotedStringInto<false>(s, buf); |
| 523 | } |
| 524 | |
| 525 | void readQuotedStringWithSQLStyle(String & s, ReadBuffer & buf) |
| 526 | { |
| 527 | s.clear(); |
| 528 | readQuotedStringInto<true>(s, buf); |
| 529 | } |
| 530 | |
| 531 | |
| 532 | template void readQuotedStringInto<true>(PaddedPODArray<UInt8> & s, ReadBuffer & buf); |
| 533 | template void readDoubleQuotedStringInto<false>(NullSink & s, ReadBuffer & buf); |
| 534 | |
| 535 | void readDoubleQuotedString(String & s, ReadBuffer & buf) |
| 536 | { |
| 537 | s.clear(); |
| 538 | readDoubleQuotedStringInto<false>(s, buf); |
| 539 | } |
| 540 | |
| 541 | void readDoubleQuotedStringWithSQLStyle(String & s, ReadBuffer & buf) |
| 542 | { |
| 543 | s.clear(); |
| 544 | readDoubleQuotedStringInto<true>(s, buf); |
| 545 | } |
| 546 | |
| 547 | void readBackQuotedString(String & s, ReadBuffer & buf) |
| 548 | { |
| 549 | s.clear(); |
| 550 | readBackQuotedStringInto<false>(s, buf); |
| 551 | } |
| 552 | |
| 553 | void readBackQuotedStringWithSQLStyle(String & s, ReadBuffer & buf) |
| 554 | { |
| 555 | s.clear(); |
| 556 | readBackQuotedStringInto<true>(s, buf); |
| 557 | } |
| 558 | |
| 559 | |
| 560 | template <typename Vector> |
| 561 | void readCSVStringInto(Vector & s, ReadBuffer & buf, const FormatSettings::CSV & settings) |
| 562 | { |
| 563 | if (buf.eof()) |
| 564 | throwReadAfterEOF(); |
| 565 | |
| 566 | const char delimiter = settings.delimiter; |
| 567 | const char maybe_quote = *buf.position(); |
| 568 | |
| 569 | /// Emptiness and not even in quotation marks. |
| 570 | if (maybe_quote == delimiter) |
| 571 | return; |
| 572 | |
| 573 | if ((settings.allow_single_quotes && maybe_quote == '\'') || (settings.allow_double_quotes && maybe_quote == '"')) |
| 574 | { |
| 575 | ++buf.position(); |
| 576 | |
| 577 | /// The quoted case. We are looking for the next quotation mark. |
| 578 | while (!buf.eof()) |
| 579 | { |
| 580 | char * next_pos = reinterpret_cast<char *>(memchr(buf.position(), maybe_quote, buf.buffer().end() - buf.position())); |
| 581 | |
| 582 | if (nullptr == next_pos) |
| 583 | next_pos = buf.buffer().end(); |
| 584 | |
| 585 | appendToStringOrVector(s, buf, next_pos); |
| 586 | buf.position() = next_pos; |
| 587 | |
| 588 | if (!buf.hasPendingData()) |
| 589 | continue; |
| 590 | |
| 591 | /// Now there is a quotation mark under the cursor. Is there any following? |
| 592 | ++buf.position(); |
| 593 | if (buf.eof()) |
| 594 | return; |
| 595 | |
| 596 | if (*buf.position() == maybe_quote) |
| 597 | { |
| 598 | s.push_back(maybe_quote); |
| 599 | ++buf.position(); |
| 600 | continue; |
| 601 | } |
| 602 | |
| 603 | return; |
| 604 | } |
| 605 | } |
| 606 | else |
| 607 | { |
| 608 | /// Unquoted case. Look for delimiter or \r or \n. |
| 609 | while (!buf.eof()) |
| 610 | { |
| 611 | char * next_pos = buf.position(); |
| 612 | |
| 613 | [&]() |
| 614 | { |
| 615 | #ifdef __SSE2__ |
| 616 | auto rc = _mm_set1_epi8('\r'); |
| 617 | auto nc = _mm_set1_epi8('\n'); |
| 618 | auto dc = _mm_set1_epi8(delimiter); |
| 619 | for (; next_pos + 15 < buf.buffer().end(); next_pos += 16) |
| 620 | { |
| 621 | __m128i bytes = _mm_loadu_si128(reinterpret_cast<const __m128i *>(next_pos)); |
| 622 | auto eq = _mm_or_si128(_mm_or_si128(_mm_cmpeq_epi8(bytes, rc), _mm_cmpeq_epi8(bytes, nc)), _mm_cmpeq_epi8(bytes, dc)); |
| 623 | uint16_t bit_mask = _mm_movemask_epi8(eq); |
| 624 | if (bit_mask) |
| 625 | { |
| 626 | next_pos += __builtin_ctz(bit_mask); |
| 627 | return; |
| 628 | } |
| 629 | } |
| 630 | #endif |
| 631 | while (next_pos < buf.buffer().end() |
| 632 | && *next_pos != delimiter && *next_pos != '\r' && *next_pos != '\n') |
| 633 | ++next_pos; |
| 634 | }(); |
| 635 | |
| 636 | |
| 637 | appendToStringOrVector(s, buf, next_pos); |
| 638 | buf.position() = next_pos; |
| 639 | |
| 640 | if (!buf.hasPendingData()) |
| 641 | continue; |
| 642 | |
| 643 | /** CSV format can contain insignificant spaces and tabs. |
| 644 | * Usually the task of skipping them is for the calling code. |
| 645 | * But in this case, it will be difficult to do this, so remove the trailing whitespace by yourself. |
| 646 | */ |
| 647 | size_t size = s.size(); |
| 648 | while (size > 0 |
| 649 | && (s[size - 1] == ' ' || s[size - 1] == '\t')) |
| 650 | --size; |
| 651 | |
| 652 | s.resize(size); |
| 653 | return; |
| 654 | } |
| 655 | } |
| 656 | } |
| 657 | |
| 658 | void readCSVString(String & s, ReadBuffer & buf, const FormatSettings::CSV & settings) |
| 659 | { |
| 660 | s.clear(); |
| 661 | readCSVStringInto(s, buf, settings); |
| 662 | } |
| 663 | |
| 664 | template void readCSVStringInto<PaddedPODArray<UInt8>>(PaddedPODArray<UInt8> & s, ReadBuffer & buf, const FormatSettings::CSV & settings); |
| 665 | |
| 666 | |
| 667 | template <typename Vector, typename ReturnType> |
| 668 | ReturnType readJSONStringInto(Vector & s, ReadBuffer & buf) |
| 669 | { |
| 670 | static constexpr bool throw_exception = std::is_same_v<ReturnType, void>; |
| 671 | |
| 672 | auto error = [](const char * message [[maybe_unused]], int code [[maybe_unused]]) |
| 673 | { |
| 674 | if constexpr (throw_exception) |
| 675 | throw Exception(message, code); |
| 676 | return ReturnType(false); |
| 677 | }; |
| 678 | |
| 679 | if (buf.eof() || *buf.position() != '"') |
| 680 | return error("Cannot parse JSON string: expected opening quote" , ErrorCodes::CANNOT_PARSE_QUOTED_STRING); |
| 681 | ++buf.position(); |
| 682 | |
| 683 | while (!buf.eof()) |
| 684 | { |
| 685 | char * next_pos = find_first_symbols<'\\', '"'>(buf.position(), buf.buffer().end()); |
| 686 | |
| 687 | appendToStringOrVector(s, buf, next_pos); |
| 688 | buf.position() = next_pos; |
| 689 | |
| 690 | if (!buf.hasPendingData()) |
| 691 | continue; |
| 692 | |
| 693 | if (*buf.position() == '"') |
| 694 | { |
| 695 | ++buf.position(); |
| 696 | return ReturnType(true); |
| 697 | } |
| 698 | |
| 699 | if (*buf.position() == '\\') |
| 700 | parseJSONEscapeSequence<Vector, ReturnType>(s, buf); |
| 701 | } |
| 702 | |
| 703 | return error("Cannot parse JSON string: expected closing quote" , ErrorCodes::CANNOT_PARSE_QUOTED_STRING); |
| 704 | } |
| 705 | |
| 706 | void readJSONString(String & s, ReadBuffer & buf) |
| 707 | { |
| 708 | s.clear(); |
| 709 | readJSONStringInto(s, buf); |
| 710 | } |
| 711 | |
| 712 | template void readJSONStringInto<PaddedPODArray<UInt8>, void>(PaddedPODArray<UInt8> & s, ReadBuffer & buf); |
| 713 | template bool readJSONStringInto<PaddedPODArray<UInt8>, bool>(PaddedPODArray<UInt8> & s, ReadBuffer & buf); |
| 714 | template void readJSONStringInto<NullSink>(NullSink & s, ReadBuffer & buf); |
| 715 | template void readJSONStringInto<String>(String & s, ReadBuffer & buf); |
| 716 | |
| 717 | |
| 718 | template <typename ReturnType> |
| 719 | ReturnType readDateTextFallback(LocalDate & date, ReadBuffer & buf) |
| 720 | { |
| 721 | static constexpr bool throw_exception = std::is_same_v<ReturnType, void>; |
| 722 | |
| 723 | auto error = [] |
| 724 | { |
| 725 | if constexpr (throw_exception) |
| 726 | throw Exception("Cannot parse date: value is too short" , ErrorCodes::CANNOT_PARSE_DATE); |
| 727 | return ReturnType(false); |
| 728 | }; |
| 729 | |
| 730 | auto ignore_delimiter = [&] |
| 731 | { |
| 732 | if (!buf.eof()) |
| 733 | { |
| 734 | ++buf.position(); |
| 735 | return true; |
| 736 | } |
| 737 | else |
| 738 | return false; |
| 739 | }; |
| 740 | |
| 741 | auto append_digit = [&](auto & x) |
| 742 | { |
| 743 | if (!buf.eof() && isNumericASCII(*buf.position())) |
| 744 | { |
| 745 | x = x * 10 + (*buf.position() - '0'); |
| 746 | ++buf.position(); |
| 747 | return true; |
| 748 | } |
| 749 | else |
| 750 | return false; |
| 751 | }; |
| 752 | |
| 753 | UInt16 year = 0; |
| 754 | if (!append_digit(year) |
| 755 | || !append_digit(year) |
| 756 | || !append_digit(year) |
| 757 | || !append_digit(year)) |
| 758 | return error(); |
| 759 | |
| 760 | if (!ignore_delimiter()) |
| 761 | return error(); |
| 762 | |
| 763 | UInt8 month = 0; |
| 764 | if (!append_digit(month)) |
| 765 | return error(); |
| 766 | append_digit(month); |
| 767 | |
| 768 | if (!ignore_delimiter()) |
| 769 | return error(); |
| 770 | |
| 771 | UInt8 day = 0; |
| 772 | if (!append_digit(day)) |
| 773 | return error(); |
| 774 | append_digit(day); |
| 775 | |
| 776 | date = LocalDate(year, month, day); |
| 777 | return ReturnType(true); |
| 778 | } |
| 779 | |
| 780 | template void readDateTextFallback<void>(LocalDate &, ReadBuffer &); |
| 781 | template bool readDateTextFallback<bool>(LocalDate &, ReadBuffer &); |
| 782 | |
| 783 | |
| 784 | template <typename ReturnType> |
| 785 | ReturnType readDateTimeTextFallback(time_t & datetime, ReadBuffer & buf, const DateLUTImpl & date_lut) |
| 786 | { |
| 787 | static constexpr bool throw_exception = std::is_same_v<ReturnType, void>; |
| 788 | |
| 789 | static constexpr auto DATE_TIME_BROKEN_DOWN_LENGTH = 19; |
| 790 | static constexpr auto UNIX_TIMESTAMP_MAX_LENGTH = 10; |
| 791 | |
| 792 | char s[DATE_TIME_BROKEN_DOWN_LENGTH]; |
| 793 | char * s_pos = s; |
| 794 | |
| 795 | /// A piece similar to unix timestamp. |
| 796 | while (s_pos < s + UNIX_TIMESTAMP_MAX_LENGTH && !buf.eof() && isNumericASCII(*buf.position())) |
| 797 | { |
| 798 | *s_pos = *buf.position(); |
| 799 | ++s_pos; |
| 800 | ++buf.position(); |
| 801 | } |
| 802 | |
| 803 | /// 2015-01-01 01:02:03 |
| 804 | if (s_pos == s + 4 && !buf.eof() && (*buf.position() < '0' || *buf.position() > '9')) |
| 805 | { |
| 806 | const size_t remaining_size = DATE_TIME_BROKEN_DOWN_LENGTH - (s_pos - s); |
| 807 | size_t size = buf.read(s_pos, remaining_size); |
| 808 | if (remaining_size != size) |
| 809 | { |
| 810 | s_pos[size] = 0; |
| 811 | |
| 812 | if constexpr (throw_exception) |
| 813 | throw Exception(std::string("Cannot parse datetime " ) + s, ErrorCodes::CANNOT_PARSE_DATETIME); |
| 814 | else |
| 815 | return false; |
| 816 | } |
| 817 | |
| 818 | UInt16 year = (s[0] - '0') * 1000 + (s[1] - '0') * 100 + (s[2] - '0') * 10 + (s[3] - '0'); |
| 819 | UInt8 month = (s[5] - '0') * 10 + (s[6] - '0'); |
| 820 | UInt8 day = (s[8] - '0') * 10 + (s[9] - '0'); |
| 821 | |
| 822 | UInt8 hour = (s[11] - '0') * 10 + (s[12] - '0'); |
| 823 | UInt8 minute = (s[14] - '0') * 10 + (s[15] - '0'); |
| 824 | UInt8 second = (s[17] - '0') * 10 + (s[18] - '0'); |
| 825 | |
| 826 | if (unlikely(year == 0)) |
| 827 | datetime = 0; |
| 828 | else |
| 829 | datetime = date_lut.makeDateTime(year, month, day, hour, minute, second); |
| 830 | } |
| 831 | else |
| 832 | { |
| 833 | /// Only unix timestamp of 5-10 characters is supported. For consistency. See readDateTimeTextImpl. |
| 834 | if (s_pos - s >= 5 && s_pos - s <= 10) |
| 835 | { |
| 836 | /// Not very efficient. |
| 837 | datetime = 0; |
| 838 | for (const char * digit_pos = s; digit_pos < s_pos; ++digit_pos) |
| 839 | datetime = datetime * 10 + *digit_pos - '0'; |
| 840 | } |
| 841 | else |
| 842 | { |
| 843 | if constexpr (throw_exception) |
| 844 | throw Exception("Cannot parse datetime" , ErrorCodes::CANNOT_PARSE_DATETIME); |
| 845 | else |
| 846 | return false; |
| 847 | } |
| 848 | } |
| 849 | |
| 850 | return ReturnType(true); |
| 851 | } |
| 852 | |
| 853 | template void readDateTimeTextFallback<void>(time_t &, ReadBuffer &, const DateLUTImpl &); |
| 854 | template bool readDateTimeTextFallback<bool>(time_t &, ReadBuffer &, const DateLUTImpl &); |
| 855 | |
| 856 | |
| 857 | void skipJSONField(ReadBuffer & buf, const StringRef & name_of_field) |
| 858 | { |
| 859 | if (buf.eof()) |
| 860 | throw Exception("Unexpected EOF for key '" + name_of_field.toString() + "'" , ErrorCodes::INCORRECT_DATA); |
| 861 | else if (*buf.position() == '"') /// skip double-quoted string |
| 862 | { |
| 863 | NullSink sink; |
| 864 | readJSONStringInto(sink, buf); |
| 865 | } |
| 866 | else if (isNumericASCII(*buf.position()) || *buf.position() == '-' || *buf.position() == '+' || *buf.position() == '.') /// skip number |
| 867 | { |
| 868 | if (*buf.position() == '+') |
| 869 | ++buf.position(); |
| 870 | |
| 871 | double v; |
| 872 | if (!tryReadFloatText(v, buf)) |
| 873 | throw Exception("Expected a number field for key '" + name_of_field.toString() + "'" , ErrorCodes::INCORRECT_DATA); |
| 874 | } |
| 875 | else if (*buf.position() == 'n') /// skip null |
| 876 | { |
| 877 | assertString("null" , buf); |
| 878 | } |
| 879 | else if (*buf.position() == 't') /// skip true |
| 880 | { |
| 881 | assertString("true" , buf); |
| 882 | } |
| 883 | else if (*buf.position() == 'f') /// skip false |
| 884 | { |
| 885 | assertString("false" , buf); |
| 886 | } |
| 887 | else if (*buf.position() == '[') |
| 888 | { |
| 889 | ++buf.position(); |
| 890 | skipWhitespaceIfAny(buf); |
| 891 | |
| 892 | if (!buf.eof() && *buf.position() == ']') /// skip empty array |
| 893 | { |
| 894 | ++buf.position(); |
| 895 | return; |
| 896 | } |
| 897 | |
| 898 | while (true) |
| 899 | { |
| 900 | skipJSONField(buf, name_of_field); |
| 901 | skipWhitespaceIfAny(buf); |
| 902 | |
| 903 | if (!buf.eof() && *buf.position() == ',') |
| 904 | { |
| 905 | ++buf.position(); |
| 906 | skipWhitespaceIfAny(buf); |
| 907 | } |
| 908 | else if (!buf.eof() && *buf.position() == ']') |
| 909 | { |
| 910 | ++buf.position(); |
| 911 | break; |
| 912 | } |
| 913 | else |
| 914 | throw Exception("Unexpected symbol for key '" + name_of_field.toString() + "'" , ErrorCodes::INCORRECT_DATA); |
| 915 | } |
| 916 | } |
| 917 | else if (*buf.position() == '{') /// skip whole object |
| 918 | { |
| 919 | ++buf.position(); |
| 920 | skipWhitespaceIfAny(buf); |
| 921 | |
| 922 | while (!buf.eof() && *buf.position() != '}') |
| 923 | { |
| 924 | // field name |
| 925 | if (*buf.position() == '"') |
| 926 | { |
| 927 | NullSink sink; |
| 928 | readJSONStringInto(sink, buf); |
| 929 | } |
| 930 | else |
| 931 | throw Exception("Unexpected symbol for key '" + name_of_field.toString() + "'" , ErrorCodes::INCORRECT_DATA); |
| 932 | |
| 933 | // ':' |
| 934 | skipWhitespaceIfAny(buf); |
| 935 | if (buf.eof() || !(*buf.position() == ':')) |
| 936 | throw Exception("Unexpected symbol for key '" + name_of_field.toString() + "'" , ErrorCodes::INCORRECT_DATA); |
| 937 | ++buf.position(); |
| 938 | skipWhitespaceIfAny(buf); |
| 939 | |
| 940 | skipJSONField(buf, name_of_field); |
| 941 | skipWhitespaceIfAny(buf); |
| 942 | |
| 943 | // optional ',' |
| 944 | if (!buf.eof() && *buf.position() == ',') |
| 945 | { |
| 946 | ++buf.position(); |
| 947 | skipWhitespaceIfAny(buf); |
| 948 | } |
| 949 | } |
| 950 | |
| 951 | if (buf.eof()) |
| 952 | throw Exception("Unexpected EOF for key '" + name_of_field.toString() + "'" , ErrorCodes::INCORRECT_DATA); |
| 953 | ++buf.position(); |
| 954 | } |
| 955 | else |
| 956 | { |
| 957 | throw Exception("Unexpected symbol '" + std::string(*buf.position(), 1) + "' for key '" + name_of_field.toString() + "'" , ErrorCodes::INCORRECT_DATA); |
| 958 | } |
| 959 | } |
| 960 | |
| 961 | |
| 962 | void readException(Exception & e, ReadBuffer & buf, const String & additional_message) |
| 963 | { |
| 964 | int code = 0; |
| 965 | String name; |
| 966 | String message; |
| 967 | String stack_trace; |
| 968 | bool has_nested = false; |
| 969 | |
| 970 | readBinary(code, buf); |
| 971 | readBinary(name, buf); |
| 972 | readBinary(message, buf); |
| 973 | readBinary(stack_trace, buf); |
| 974 | readBinary(has_nested, buf); |
| 975 | |
| 976 | WriteBufferFromOwnString out; |
| 977 | |
| 978 | if (!additional_message.empty()) |
| 979 | out << additional_message << ". " ; |
| 980 | |
| 981 | if (name != "DB::Exception" ) |
| 982 | out << name << ". " ; |
| 983 | |
| 984 | out << message << "." ; |
| 985 | |
| 986 | if (!stack_trace.empty()) |
| 987 | out << " Stack trace:\n\n" << stack_trace; |
| 988 | |
| 989 | if (has_nested) |
| 990 | { |
| 991 | Exception nested; |
| 992 | readException(nested, buf); |
| 993 | e = Exception(out.str(), nested, code); |
| 994 | } |
| 995 | else |
| 996 | e = Exception(out.str(), code); |
| 997 | } |
| 998 | |
| 999 | void readAndThrowException(ReadBuffer & buf, const String & additional_message) |
| 1000 | { |
| 1001 | Exception e; |
| 1002 | readException(e, buf, additional_message); |
| 1003 | e.rethrow(); |
| 1004 | } |
| 1005 | |
| 1006 | |
| 1007 | void skipToNextLineOrEOF(ReadBuffer & buf) |
| 1008 | { |
| 1009 | while (!buf.eof()) |
| 1010 | { |
| 1011 | char * next_pos = find_first_symbols<'\n'>(buf.position(), buf.buffer().end()); |
| 1012 | buf.position() = next_pos; |
| 1013 | |
| 1014 | if (!buf.hasPendingData()) |
| 1015 | continue; |
| 1016 | |
| 1017 | if (*buf.position() == '\n') |
| 1018 | { |
| 1019 | ++buf.position(); |
| 1020 | return; |
| 1021 | } |
| 1022 | } |
| 1023 | } |
| 1024 | |
| 1025 | |
| 1026 | void skipToUnescapedNextLineOrEOF(ReadBuffer & buf) |
| 1027 | { |
| 1028 | while (!buf.eof()) |
| 1029 | { |
| 1030 | char * next_pos = find_first_symbols<'\n', '\\'>(buf.position(), buf.buffer().end()); |
| 1031 | buf.position() = next_pos; |
| 1032 | |
| 1033 | if (!buf.hasPendingData()) |
| 1034 | continue; |
| 1035 | |
| 1036 | if (*buf.position() == '\n') |
| 1037 | { |
| 1038 | ++buf.position(); |
| 1039 | return; |
| 1040 | } |
| 1041 | |
| 1042 | if (*buf.position() == '\\') |
| 1043 | { |
| 1044 | ++buf.position(); |
| 1045 | if (buf.eof()) |
| 1046 | return; |
| 1047 | |
| 1048 | /// Skip escaped character. We do not consider escape sequences with more than one character after backslash (\x01). |
| 1049 | /// It's ok for the purpose of this function, because we are interested only in \n and \\. |
| 1050 | ++buf.position(); |
| 1051 | continue; |
| 1052 | } |
| 1053 | } |
| 1054 | } |
| 1055 | |
| 1056 | void saveUpToPosition(ReadBuffer & in, DB::Memory<> & memory, char * current) |
| 1057 | { |
| 1058 | assert(current >= in.position()); |
| 1059 | assert(current <= in.buffer().end()); |
| 1060 | |
| 1061 | const int old_bytes = memory.size(); |
| 1062 | const int additional_bytes = current - in.position(); |
| 1063 | const int new_bytes = old_bytes + additional_bytes; |
| 1064 | /// There are no new bytes to add to memory. |
| 1065 | /// No need to do extra stuff. |
| 1066 | if (new_bytes == 0) |
| 1067 | return; |
| 1068 | memory.resize(new_bytes); |
| 1069 | memcpy(memory.data() + old_bytes, in.position(), additional_bytes); |
| 1070 | in.position() = current; |
| 1071 | } |
| 1072 | |
| 1073 | bool loadAtPosition(ReadBuffer & in, DB::Memory<> & memory, char * & current) |
| 1074 | { |
| 1075 | assert(current <= in.buffer().end()); |
| 1076 | |
| 1077 | if (current < in.buffer().end()) |
| 1078 | return true; |
| 1079 | |
| 1080 | saveUpToPosition(in, memory, current); |
| 1081 | bool loaded_more = !in.eof(); |
| 1082 | assert(in.position() == in.buffer().begin()); |
| 1083 | current = in.position(); |
| 1084 | return loaded_more; |
| 1085 | } |
| 1086 | |
| 1087 | } |
| 1088 | |