| 1 | #include <Storages/StorageLog.h> |
| 2 | #include <Storages/StorageFactory.h> |
| 3 | |
| 4 | #include <Common/Exception.h> |
| 5 | #include <Common/StringUtils/StringUtils.h> |
| 6 | |
| 7 | #include <IO/ReadBufferFromFile.h> |
| 8 | #include <IO/WriteBufferFromFile.h> |
| 9 | #include <Compression/CompressedReadBuffer.h> |
| 10 | #include <Compression/CompressedWriteBuffer.h> |
| 11 | #include <IO/ReadHelpers.h> |
| 12 | #include <IO/WriteHelpers.h> |
| 13 | |
| 14 | #include <DataTypes/NestedUtils.h> |
| 15 | |
| 16 | #include <DataStreams/IBlockInputStream.h> |
| 17 | #include <DataStreams/IBlockOutputStream.h> |
| 18 | |
| 19 | #include <Columns/ColumnArray.h> |
| 20 | |
| 21 | #include <Common/typeid_cast.h> |
| 22 | |
| 23 | #include <Interpreters/Context.h> |
| 24 | |
| 25 | #include <Poco/Path.h> |
| 26 | #include <Poco/DirectoryIterator.h> |
| 27 | |
| 28 | |
| 29 | #define DBMS_STORAGE_LOG_DATA_FILE_EXTENSION ".bin" |
| 30 | #define DBMS_STORAGE_LOG_MARKS_FILE_NAME "__marks.mrk" |
| 31 | |
| 32 | |
| 33 | namespace DB |
| 34 | { |
| 35 | |
| 36 | namespace ErrorCodes |
| 37 | { |
| 38 | extern const int LOGICAL_ERROR; |
| 39 | extern const int EMPTY_LIST_OF_COLUMNS_PASSED; |
| 40 | extern const int NO_SUCH_COLUMN_IN_TABLE; |
| 41 | extern const int DUPLICATE_COLUMN; |
| 42 | extern const int SIZES_OF_MARKS_FILES_ARE_INCONSISTENT; |
| 43 | extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH; |
| 44 | extern const int INCORRECT_FILE_NAME; |
| 45 | } |
| 46 | |
| 47 | |
| 48 | class LogBlockInputStream final : public IBlockInputStream |
| 49 | { |
| 50 | public: |
| 51 | LogBlockInputStream( |
| 52 | size_t block_size_, const NamesAndTypesList & columns_, StorageLog & storage_, |
| 53 | size_t mark_number_, size_t rows_limit_, size_t max_read_buffer_size_) |
| 54 | : block_size(block_size_), |
| 55 | columns(columns_), |
| 56 | storage(storage_), |
| 57 | mark_number(mark_number_), |
| 58 | rows_limit(rows_limit_), |
| 59 | max_read_buffer_size(max_read_buffer_size_) |
| 60 | { |
| 61 | } |
| 62 | |
| 63 | String getName() const override { return "Log" ; } |
| 64 | |
| 65 | Block () const override |
| 66 | { |
| 67 | Block res; |
| 68 | |
| 69 | for (const auto & name_type : columns) |
| 70 | res.insert({ name_type.type->createColumn(), name_type.type, name_type.name }); |
| 71 | |
| 72 | return Nested::flatten(res); |
| 73 | } |
| 74 | |
| 75 | protected: |
| 76 | Block readImpl() override; |
| 77 | |
| 78 | private: |
| 79 | |
| 80 | size_t block_size; |
| 81 | NamesAndTypesList columns; |
| 82 | StorageLog & storage; |
| 83 | size_t mark_number; /// from what mark to read data |
| 84 | size_t rows_limit; /// The maximum number of rows that can be read |
| 85 | size_t rows_read = 0; |
| 86 | size_t max_read_buffer_size; |
| 87 | |
| 88 | struct Stream |
| 89 | { |
| 90 | Stream(const std::string & data_path, size_t offset, size_t max_read_buffer_size_) |
| 91 | : plain(data_path, std::min(static_cast<Poco::File::FileSize>(max_read_buffer_size_), Poco::File(data_path).getSize())), |
| 92 | compressed(plain) |
| 93 | { |
| 94 | if (offset) |
| 95 | plain.seek(offset); |
| 96 | } |
| 97 | |
| 98 | ReadBufferFromFile plain; |
| 99 | CompressedReadBuffer compressed; |
| 100 | }; |
| 101 | |
| 102 | using FileStreams = std::map<std::string, Stream>; |
| 103 | FileStreams streams; |
| 104 | |
| 105 | using DeserializeState = IDataType::DeserializeBinaryBulkStatePtr; |
| 106 | using DeserializeStates = std::map<String, DeserializeState>; |
| 107 | DeserializeStates deserialize_states; |
| 108 | |
| 109 | void readData(const String & name, const IDataType & type, IColumn & column, size_t max_rows_to_read); |
| 110 | |
| 111 | }; |
| 112 | |
| 113 | |
| 114 | class LogBlockOutputStream final : public IBlockOutputStream |
| 115 | { |
| 116 | public: |
| 117 | explicit LogBlockOutputStream(StorageLog & storage_) |
| 118 | : storage(storage_), |
| 119 | lock(storage.rwlock), |
| 120 | marks_stream(storage.marks_file.path(), 4096, O_APPEND | O_CREAT | O_WRONLY) |
| 121 | { |
| 122 | } |
| 123 | |
| 124 | ~LogBlockOutputStream() override |
| 125 | { |
| 126 | try |
| 127 | { |
| 128 | writeSuffix(); |
| 129 | } |
| 130 | catch (...) |
| 131 | { |
| 132 | tryLogCurrentException(__PRETTY_FUNCTION__); |
| 133 | } |
| 134 | } |
| 135 | |
| 136 | Block () const override { return storage.getSampleBlock(); } |
| 137 | void write(const Block & block) override; |
| 138 | void writeSuffix() override; |
| 139 | |
| 140 | private: |
| 141 | StorageLog & storage; |
| 142 | std::unique_lock<std::shared_mutex> lock; |
| 143 | bool done = false; |
| 144 | |
| 145 | struct Stream |
| 146 | { |
| 147 | Stream(const std::string & data_path, CompressionCodecPtr codec, size_t max_compress_block_size) : |
| 148 | plain(data_path, max_compress_block_size, O_APPEND | O_CREAT | O_WRONLY), |
| 149 | compressed(plain, std::move(codec), max_compress_block_size) |
| 150 | { |
| 151 | plain_offset = Poco::File(data_path).getSize(); |
| 152 | } |
| 153 | |
| 154 | WriteBufferFromFile plain; |
| 155 | CompressedWriteBuffer compressed; |
| 156 | |
| 157 | size_t plain_offset; /// How many bytes were in the file at the time the LogBlockOutputStream was created. |
| 158 | |
| 159 | void finalize() |
| 160 | { |
| 161 | compressed.next(); |
| 162 | plain.next(); |
| 163 | } |
| 164 | }; |
| 165 | |
| 166 | using Mark = StorageLog::Mark; |
| 167 | using MarksForColumns = std::vector<std::pair<size_t, Mark>>; |
| 168 | |
| 169 | using FileStreams = std::map<std::string, Stream>; |
| 170 | FileStreams streams; |
| 171 | |
| 172 | using WrittenStreams = std::set<std::string>; |
| 173 | |
| 174 | WriteBufferFromFile marks_stream; /// Declared below `lock` to make the file open when rwlock is captured. |
| 175 | |
| 176 | using SerializeState = IDataType::SerializeBinaryBulkStatePtr; |
| 177 | using SerializeStates = std::map<String, SerializeState>; |
| 178 | SerializeStates serialize_states; |
| 179 | |
| 180 | IDataType::OutputStreamGetter createStreamGetter(const String & name, WrittenStreams & written_streams); |
| 181 | |
| 182 | void writeData(const String & name, const IDataType & type, const IColumn & column, |
| 183 | MarksForColumns & out_marks, |
| 184 | WrittenStreams & written_streams); |
| 185 | |
| 186 | void writeMarks(MarksForColumns && marks); |
| 187 | }; |
| 188 | |
| 189 | |
| 190 | Block LogBlockInputStream::readImpl() |
| 191 | { |
| 192 | Block res; |
| 193 | |
| 194 | if (rows_read == rows_limit) |
| 195 | return res; |
| 196 | |
| 197 | /// If there are no files in the folder, the table is empty. |
| 198 | if (Poco::DirectoryIterator(storage.getFullPath()) == Poco::DirectoryIterator()) |
| 199 | return res; |
| 200 | |
| 201 | /// How many rows to read for the next block. |
| 202 | size_t max_rows_to_read = std::min(block_size, rows_limit - rows_read); |
| 203 | |
| 204 | for (const auto & name_type : columns) |
| 205 | { |
| 206 | MutableColumnPtr column = name_type.type->createColumn(); |
| 207 | |
| 208 | try |
| 209 | { |
| 210 | readData(name_type.name, *name_type.type, *column, max_rows_to_read); |
| 211 | } |
| 212 | catch (Exception & e) |
| 213 | { |
| 214 | e.addMessage("while reading column " + name_type.name + " at " + storage.path); |
| 215 | throw; |
| 216 | } |
| 217 | |
| 218 | if (column->size()) |
| 219 | res.insert(ColumnWithTypeAndName(std::move(column), name_type.type, name_type.name)); |
| 220 | } |
| 221 | |
| 222 | if (res) |
| 223 | rows_read += res.rows(); |
| 224 | |
| 225 | if (!res || rows_read == rows_limit) |
| 226 | { |
| 227 | /** Close the files (before destroying the object). |
| 228 | * When many sources are created, but simultaneously reading only a few of them, |
| 229 | * buffers don't waste memory. |
| 230 | */ |
| 231 | streams.clear(); |
| 232 | } |
| 233 | |
| 234 | return Nested::flatten(res); |
| 235 | } |
| 236 | |
| 237 | |
| 238 | void LogBlockInputStream::readData(const String & name, const IDataType & type, IColumn & column, size_t max_rows_to_read) |
| 239 | { |
| 240 | IDataType::DeserializeBinaryBulkSettings settings; /// TODO Use avg_value_size_hint. |
| 241 | |
| 242 | auto createStringGetter = [&](bool stream_for_prefix) |
| 243 | { |
| 244 | return [&, stream_for_prefix] (const IDataType::SubstreamPath & path) -> ReadBuffer * |
| 245 | { |
| 246 | String stream_name = IDataType::getFileNameForStream(name, path); |
| 247 | |
| 248 | const auto & file_it = storage.files.find(stream_name); |
| 249 | if (storage.files.end() == file_it) |
| 250 | throw Exception("Logical error: no information about file " + stream_name + " in StorageLog" , ErrorCodes::LOGICAL_ERROR); |
| 251 | |
| 252 | UInt64 offset = 0; |
| 253 | if (!stream_for_prefix && mark_number) |
| 254 | offset = file_it->second.marks[mark_number].offset; |
| 255 | |
| 256 | auto & data_file_path = file_it->second.data_file.path(); |
| 257 | auto it = streams.try_emplace(stream_name, data_file_path, offset, max_read_buffer_size).first; |
| 258 | return &it->second.compressed; |
| 259 | }; |
| 260 | }; |
| 261 | |
| 262 | if (deserialize_states.count(name) == 0) |
| 263 | { |
| 264 | settings.getter = createStringGetter(true); |
| 265 | type.deserializeBinaryBulkStatePrefix(settings, deserialize_states[name]); |
| 266 | } |
| 267 | |
| 268 | settings.getter = createStringGetter(false); |
| 269 | type.deserializeBinaryBulkWithMultipleStreams(column, max_rows_to_read, settings, deserialize_states[name]); |
| 270 | } |
| 271 | |
| 272 | |
| 273 | void LogBlockOutputStream::write(const Block & block) |
| 274 | { |
| 275 | storage.check(block, true); |
| 276 | |
| 277 | /// The set of written offset columns so that you do not write shared offsets of columns for nested structures multiple times |
| 278 | WrittenStreams written_streams; |
| 279 | |
| 280 | MarksForColumns marks; |
| 281 | marks.reserve(storage.file_count); |
| 282 | |
| 283 | for (size_t i = 0; i < block.columns(); ++i) |
| 284 | { |
| 285 | const ColumnWithTypeAndName & column = block.safeGetByPosition(i); |
| 286 | writeData(column.name, *column.type, *column.column, marks, written_streams); |
| 287 | } |
| 288 | |
| 289 | writeMarks(std::move(marks)); |
| 290 | } |
| 291 | |
| 292 | |
| 293 | void LogBlockOutputStream::writeSuffix() |
| 294 | { |
| 295 | if (done) |
| 296 | return; |
| 297 | done = true; |
| 298 | |
| 299 | WrittenStreams written_streams; |
| 300 | IDataType::SerializeBinaryBulkSettings settings; |
| 301 | for (const auto & column : getHeader()) |
| 302 | { |
| 303 | auto it = serialize_states.find(column.name); |
| 304 | if (it != serialize_states.end()) |
| 305 | { |
| 306 | settings.getter = createStreamGetter(column.name, written_streams); |
| 307 | column.type->serializeBinaryBulkStateSuffix(settings, it->second); |
| 308 | } |
| 309 | } |
| 310 | |
| 311 | /// Finish write. |
| 312 | marks_stream.next(); |
| 313 | |
| 314 | for (auto & name_stream : streams) |
| 315 | name_stream.second.finalize(); |
| 316 | |
| 317 | std::vector<Poco::File> column_files; |
| 318 | for (const auto & name_stream : streams) |
| 319 | column_files.push_back(storage.files[name_stream.first].data_file); |
| 320 | column_files.push_back(storage.marks_file); |
| 321 | |
| 322 | storage.file_checker.update(column_files.begin(), column_files.end()); |
| 323 | |
| 324 | streams.clear(); |
| 325 | } |
| 326 | |
| 327 | |
| 328 | IDataType::OutputStreamGetter LogBlockOutputStream::createStreamGetter(const String & name, |
| 329 | WrittenStreams & written_streams) |
| 330 | { |
| 331 | return [&] (const IDataType::SubstreamPath & path) -> WriteBuffer * |
| 332 | { |
| 333 | String stream_name = IDataType::getFileNameForStream(name, path); |
| 334 | if (written_streams.count(stream_name)) |
| 335 | return nullptr; |
| 336 | |
| 337 | auto it = streams.find(stream_name); |
| 338 | if (streams.end() == it) |
| 339 | throw Exception("Logical error: stream was not created when writing data in LogBlockOutputStream" , |
| 340 | ErrorCodes::LOGICAL_ERROR); |
| 341 | return &it->second.compressed; |
| 342 | }; |
| 343 | } |
| 344 | |
| 345 | |
| 346 | void LogBlockOutputStream::writeData(const String & name, const IDataType & type, const IColumn & column, |
| 347 | MarksForColumns & out_marks, |
| 348 | WrittenStreams & written_streams) |
| 349 | { |
| 350 | IDataType::SerializeBinaryBulkSettings settings; |
| 351 | |
| 352 | type.enumerateStreams([&] (const IDataType::SubstreamPath & path) |
| 353 | { |
| 354 | String stream_name = IDataType::getFileNameForStream(name, path); |
| 355 | if (written_streams.count(stream_name)) |
| 356 | return; |
| 357 | |
| 358 | const auto & columns = storage.getColumns(); |
| 359 | streams.try_emplace( |
| 360 | stream_name, |
| 361 | storage.files[stream_name].data_file.path(), |
| 362 | columns.getCodecOrDefault(name), |
| 363 | storage.max_compress_block_size); |
| 364 | }, settings.path); |
| 365 | |
| 366 | settings.getter = createStreamGetter(name, written_streams); |
| 367 | |
| 368 | if (serialize_states.count(name) == 0) |
| 369 | type.serializeBinaryBulkStatePrefix(settings, serialize_states[name]); |
| 370 | |
| 371 | type.enumerateStreams([&] (const IDataType::SubstreamPath & path) |
| 372 | { |
| 373 | String stream_name = IDataType::getFileNameForStream(name, path); |
| 374 | if (written_streams.count(stream_name)) |
| 375 | return; |
| 376 | |
| 377 | const auto & file = storage.files[stream_name]; |
| 378 | const auto stream_it = streams.find(stream_name); |
| 379 | |
| 380 | Mark mark; |
| 381 | mark.rows = (file.marks.empty() ? 0 : file.marks.back().rows) + column.size(); |
| 382 | mark.offset = stream_it->second.plain_offset + stream_it->second.plain.count(); |
| 383 | |
| 384 | out_marks.emplace_back(file.column_index, mark); |
| 385 | }, settings.path); |
| 386 | |
| 387 | type.serializeBinaryBulkWithMultipleStreams(column, 0, 0, settings, serialize_states[name]); |
| 388 | |
| 389 | type.enumerateStreams([&] (const IDataType::SubstreamPath & path) |
| 390 | { |
| 391 | String stream_name = IDataType::getFileNameForStream(name, path); |
| 392 | if (!written_streams.emplace(stream_name).second) |
| 393 | return; |
| 394 | |
| 395 | auto it = streams.find(stream_name); |
| 396 | if (streams.end() == it) |
| 397 | throw Exception("Logical error: stream was not created when writing data in LogBlockOutputStream" , ErrorCodes::LOGICAL_ERROR); |
| 398 | it->second.compressed.next(); |
| 399 | }, settings.path); |
| 400 | } |
| 401 | |
| 402 | |
| 403 | void LogBlockOutputStream::writeMarks(MarksForColumns && marks) |
| 404 | { |
| 405 | if (marks.size() != storage.file_count) |
| 406 | throw Exception("Wrong number of marks generated from block. Makes no sense." , ErrorCodes::LOGICAL_ERROR); |
| 407 | |
| 408 | std::sort(marks.begin(), marks.end(), [](const auto & a, const auto & b) { return a.first < b.first; }); |
| 409 | |
| 410 | for (const auto & mark : marks) |
| 411 | { |
| 412 | writeIntBinary(mark.second.rows, marks_stream); |
| 413 | writeIntBinary(mark.second.offset, marks_stream); |
| 414 | |
| 415 | size_t column_index = mark.first; |
| 416 | storage.files[storage.column_names_by_idx[column_index]].marks.push_back(mark.second); |
| 417 | } |
| 418 | } |
| 419 | |
| 420 | StorageLog::StorageLog( |
| 421 | const std::string & relative_path_, |
| 422 | const std::string & database_name_, |
| 423 | const std::string & table_name_, |
| 424 | const ColumnsDescription & columns_, |
| 425 | const ConstraintsDescription & constraints_, |
| 426 | size_t max_compress_block_size_, |
| 427 | const Context & context_) |
| 428 | : base_path(context_.getPath()), path(base_path + relative_path_), table_name(table_name_), database_name(database_name_), |
| 429 | max_compress_block_size(max_compress_block_size_), |
| 430 | file_checker(path + "sizes.json" ) |
| 431 | { |
| 432 | setColumns(columns_); |
| 433 | setConstraints(constraints_); |
| 434 | |
| 435 | if (relative_path_.empty()) |
| 436 | throw Exception("Storage " + getName() + " requires data path" , ErrorCodes::INCORRECT_FILE_NAME); |
| 437 | |
| 438 | /// create files if they do not exist |
| 439 | Poco::File(path).createDirectories(); |
| 440 | |
| 441 | for (const auto & column : getColumns().getAllPhysical()) |
| 442 | addFiles(column.name, *column.type); |
| 443 | |
| 444 | marks_file = Poco::File(path + DBMS_STORAGE_LOG_MARKS_FILE_NAME); |
| 445 | } |
| 446 | |
| 447 | |
| 448 | void StorageLog::addFiles(const String & column_name, const IDataType & type) |
| 449 | { |
| 450 | if (files.end() != files.find(column_name)) |
| 451 | throw Exception("Duplicate column with name " + column_name + " in constructor of StorageLog." , |
| 452 | ErrorCodes::DUPLICATE_COLUMN); |
| 453 | |
| 454 | IDataType::StreamCallback stream_callback = [&] (const IDataType::SubstreamPath & substream_path) |
| 455 | { |
| 456 | String stream_name = IDataType::getFileNameForStream(column_name, substream_path); |
| 457 | |
| 458 | if (!files.count(stream_name)) |
| 459 | { |
| 460 | ColumnData & column_data = files[stream_name]; |
| 461 | column_data.column_index = file_count; |
| 462 | column_data.data_file = Poco::File{ |
| 463 | path + stream_name + DBMS_STORAGE_LOG_DATA_FILE_EXTENSION}; |
| 464 | |
| 465 | column_names_by_idx.push_back(stream_name); |
| 466 | ++file_count; |
| 467 | } |
| 468 | }; |
| 469 | |
| 470 | IDataType::SubstreamPath substream_path; |
| 471 | type.enumerateStreams(stream_callback, substream_path); |
| 472 | } |
| 473 | |
| 474 | |
| 475 | void StorageLog::loadMarks() |
| 476 | { |
| 477 | std::unique_lock<std::shared_mutex> lock(rwlock); |
| 478 | |
| 479 | if (loaded_marks) |
| 480 | return; |
| 481 | |
| 482 | using FilesByIndex = std::vector<Files_t::iterator>; |
| 483 | |
| 484 | FilesByIndex files_by_index(file_count); |
| 485 | for (Files_t::iterator it = files.begin(); it != files.end(); ++it) |
| 486 | files_by_index[it->second.column_index] = it; |
| 487 | |
| 488 | if (marks_file.exists()) |
| 489 | { |
| 490 | size_t file_size = marks_file.getSize(); |
| 491 | if (file_size % (file_count * sizeof(Mark)) != 0) |
| 492 | throw Exception("Size of marks file is inconsistent" , ErrorCodes::SIZES_OF_MARKS_FILES_ARE_INCONSISTENT); |
| 493 | |
| 494 | size_t marks_count = file_size / (file_count * sizeof(Mark)); |
| 495 | |
| 496 | for (auto & file : files_by_index) |
| 497 | file->second.marks.reserve(marks_count); |
| 498 | |
| 499 | ReadBufferFromFile marks_rb(marks_file.path(), 32768); |
| 500 | while (!marks_rb.eof()) |
| 501 | { |
| 502 | for (size_t i = 0; i < files_by_index.size(); ++i) |
| 503 | { |
| 504 | Mark mark; |
| 505 | readIntBinary(mark.rows, marks_rb); |
| 506 | readIntBinary(mark.offset, marks_rb); |
| 507 | files_by_index[i]->second.marks.push_back(mark); |
| 508 | } |
| 509 | } |
| 510 | } |
| 511 | |
| 512 | loaded_marks = true; |
| 513 | } |
| 514 | |
| 515 | |
| 516 | void StorageLog::rename(const String & new_path_to_table_data, const String & new_database_name, const String & new_table_name, TableStructureWriteLockHolder &) |
| 517 | { |
| 518 | std::unique_lock<std::shared_mutex> lock(rwlock); |
| 519 | |
| 520 | /// Rename directory with data. |
| 521 | String new_path = base_path + new_path_to_table_data; |
| 522 | Poco::File(path).renameTo(new_path); |
| 523 | |
| 524 | path = new_path; |
| 525 | table_name = new_table_name; |
| 526 | database_name = new_database_name; |
| 527 | file_checker.setPath(path + "sizes.json" ); |
| 528 | |
| 529 | for (auto & file : files) |
| 530 | file.second.data_file = Poco::File(path + Poco::Path(file.second.data_file.path()).getFileName()); |
| 531 | |
| 532 | marks_file = Poco::File(path + DBMS_STORAGE_LOG_MARKS_FILE_NAME); |
| 533 | } |
| 534 | |
| 535 | void StorageLog::truncate(const ASTPtr &, const Context &, TableStructureWriteLockHolder &) |
| 536 | { |
| 537 | std::shared_lock<std::shared_mutex> lock(rwlock); |
| 538 | |
| 539 | String table_dir = path; |
| 540 | |
| 541 | files.clear(); |
| 542 | file_count = 0; |
| 543 | loaded_marks = false; |
| 544 | |
| 545 | std::vector<Poco::File> data_files; |
| 546 | Poco::File(table_dir).list(data_files); |
| 547 | |
| 548 | for (auto & file : data_files) |
| 549 | file.remove(false); |
| 550 | |
| 551 | for (const auto & column : getColumns().getAllPhysical()) |
| 552 | addFiles(column.name, *column.type); |
| 553 | |
| 554 | file_checker = FileChecker{table_dir + "/" + "sizes.json" }; |
| 555 | marks_file = Poco::File(table_dir + "/" + DBMS_STORAGE_LOG_MARKS_FILE_NAME); |
| 556 | } |
| 557 | |
| 558 | |
| 559 | const StorageLog::Marks & StorageLog::getMarksWithRealRowCount() const |
| 560 | { |
| 561 | const String & column_name = getColumns().begin()->name; |
| 562 | const IDataType & column_type = *getColumns().begin()->type; |
| 563 | String filename; |
| 564 | |
| 565 | /** We take marks from first column. |
| 566 | * If this is a data type with multiple stream, get the first stream, that we assume have real row count. |
| 567 | * (Example: for Array data type, first stream is array sizes; and number of array sizes is the number of arrays). |
| 568 | */ |
| 569 | IDataType::SubstreamPath substream_root_path; |
| 570 | column_type.enumerateStreams([&](const IDataType::SubstreamPath & substream_path) |
| 571 | { |
| 572 | if (filename.empty()) |
| 573 | filename = IDataType::getFileNameForStream(column_name, substream_path); |
| 574 | }, substream_root_path); |
| 575 | |
| 576 | Files_t::const_iterator it = files.find(filename); |
| 577 | if (files.end() == it) |
| 578 | throw Exception("Cannot find file " + filename, ErrorCodes::LOGICAL_ERROR); |
| 579 | |
| 580 | return it->second.marks; |
| 581 | } |
| 582 | |
| 583 | BlockInputStreams StorageLog::read( |
| 584 | const Names & column_names, |
| 585 | const SelectQueryInfo & /*query_info*/, |
| 586 | const Context & context, |
| 587 | QueryProcessingStage::Enum /*processed_stage*/, |
| 588 | size_t max_block_size, |
| 589 | unsigned num_streams) |
| 590 | { |
| 591 | check(column_names); |
| 592 | loadMarks(); |
| 593 | |
| 594 | NamesAndTypesList all_columns = Nested::collect(getColumns().getAllPhysical().addTypes(column_names)); |
| 595 | |
| 596 | std::shared_lock<std::shared_mutex> lock(rwlock); |
| 597 | |
| 598 | BlockInputStreams res; |
| 599 | |
| 600 | const Marks & marks = getMarksWithRealRowCount(); |
| 601 | size_t marks_size = marks.size(); |
| 602 | |
| 603 | if (num_streams > marks_size) |
| 604 | num_streams = marks_size; |
| 605 | |
| 606 | size_t max_read_buffer_size = context.getSettingsRef().max_read_buffer_size; |
| 607 | |
| 608 | for (size_t stream = 0; stream < num_streams; ++stream) |
| 609 | { |
| 610 | size_t mark_begin = stream * marks_size / num_streams; |
| 611 | size_t mark_end = (stream + 1) * marks_size / num_streams; |
| 612 | |
| 613 | size_t rows_begin = mark_begin ? marks[mark_begin - 1].rows : 0; |
| 614 | size_t rows_end = mark_end ? marks[mark_end - 1].rows : 0; |
| 615 | |
| 616 | res.emplace_back(std::make_shared<LogBlockInputStream>( |
| 617 | max_block_size, |
| 618 | all_columns, |
| 619 | *this, |
| 620 | mark_begin, |
| 621 | rows_end - rows_begin, |
| 622 | max_read_buffer_size)); |
| 623 | } |
| 624 | |
| 625 | return res; |
| 626 | } |
| 627 | |
| 628 | BlockOutputStreamPtr StorageLog::write( |
| 629 | const ASTPtr & /*query*/, const Context & /*context*/) |
| 630 | { |
| 631 | loadMarks(); |
| 632 | return std::make_shared<LogBlockOutputStream>(*this); |
| 633 | } |
| 634 | |
| 635 | CheckResults StorageLog::checkData(const ASTPtr & /* query */, const Context & /* context */) |
| 636 | { |
| 637 | std::shared_lock<std::shared_mutex> lock(rwlock); |
| 638 | return file_checker.check(); |
| 639 | } |
| 640 | |
| 641 | |
| 642 | void registerStorageLog(StorageFactory & factory) |
| 643 | { |
| 644 | factory.registerStorage("Log" , [](const StorageFactory::Arguments & args) |
| 645 | { |
| 646 | if (!args.engine_args.empty()) |
| 647 | throw Exception( |
| 648 | "Engine " + args.engine_name + " doesn't support any arguments (" + toString(args.engine_args.size()) + " given)" , |
| 649 | ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH); |
| 650 | |
| 651 | return StorageLog::create( |
| 652 | args.relative_data_path, args.database_name, args.table_name, args.columns, args.constraints, |
| 653 | args.context.getSettings().max_compress_block_size, args.context); |
| 654 | }); |
| 655 | } |
| 656 | |
| 657 | } |
| 658 | |