#include #include #include #include #include #include #include #include #include #include namespace ProfileEvents { extern const Event ExternalJoinWritePart; extern const Event ExternalJoinMerge; extern const Event ExternalJoinCompressedBytes; extern const Event ExternalJoinUncompressedBytes; extern const Event ExternalProcessingCompressedBytesTotal; extern const Event ExternalProcessingUncompressedBytesTotal; } namespace CurrentMetrics { extern const Metric TemporaryFilesForJoin; } namespace DB { namespace { TemporaryFileOnDiskHolder flushToFile(const DiskPtr & disk, const Block & header, QueryPipelineBuilder pipeline, const String & codec) { auto tmp_file = std::make_unique(disk, CurrentMetrics::TemporaryFilesForJoin); auto write_stat = TemporaryFileStream::write(tmp_file->getPath(), header, std::move(pipeline), codec); ProfileEvents::increment(ProfileEvents::ExternalProcessingCompressedBytesTotal, write_stat.compressed_bytes); ProfileEvents::increment(ProfileEvents::ExternalProcessingUncompressedBytesTotal, write_stat.uncompressed_bytes); ProfileEvents::increment(ProfileEvents::ExternalJoinCompressedBytes, write_stat.compressed_bytes); ProfileEvents::increment(ProfileEvents::ExternalJoinUncompressedBytes, write_stat.uncompressed_bytes); ProfileEvents::increment(ProfileEvents::ExternalJoinWritePart); return tmp_file; } SortedBlocksWriter::SortedFiles flushToManyFiles(const DiskPtr & disk, const Block & header, QueryPipelineBuilder builder, const String & codec, std::function callback = [](const Block &){}) { std::vector files; auto pipeline = QueryPipelineBuilder::getPipeline(std::move(builder)); PullingPipelineExecutor executor(pipeline); Block block; while (executor.pull(block)) { if (!block.rows()) continue; callback(block); QueryPipelineBuilder one_block_pipeline; Chunk chunk(block.getColumns(), block.rows()); one_block_pipeline.init(Pipe(std::make_shared(block.cloneEmpty(), std::move(chunk)))); auto tmp_file = flushToFile(disk, header, std::move(one_block_pipeline), codec); files.emplace_back(std::move(tmp_file)); } return files; } } void SortedBlocksWriter::insert(Block && block) { bool can_insert_more = false; bool has_data_to_flush = false; BlocksList current_blocks; size_t row_count = 0; size_t bytes = 0; size_t flush_no = 0; { std::lock_guard lock{insert_mutex}; /// insert bock into BlocksList undef lock inserted_blocks.insert(std::move(block)); size_t total_row_count = inserted_blocks.row_count + row_count_in_flush; size_t total_bytes = inserted_blocks.bytes + bytes_in_flush; can_insert_more = size_limits.softCheck(total_row_count, total_bytes); has_data_to_flush = !size_limits.softCheck(inserted_blocks.row_count * num_streams, inserted_blocks.bytes * num_streams); if (has_data_to_flush) { ++flush_inflight; current_blocks.swap(inserted_blocks.blocks); row_count_in_flush = total_row_count; bytes_in_flush = total_bytes; row_count = inserted_blocks.row_count; bytes = inserted_blocks.bytes; inserted_blocks.clear(); } else if (can_insert_more) flush_no = flush_number; } if (has_data_to_flush) { /// flush new blocks without lock auto flushed = flush(current_blocks); current_blocks.clear(); std::lock_guard lock{insert_mutex}; sorted_files.emplace_back(std::move(flushed)); row_count_in_flush -= row_count; bytes_in_flush -= bytes; /// notify another insert (flush_number) and merge (flush_inflight) ++flush_number; --flush_inflight; flush_condvar.notify_all(); } else if (!can_insert_more) { /// wakeup insert blocked by out of limit std::unique_lock lock{insert_mutex}; flush_condvar.wait(lock, [&]{ return flush_no < flush_number; }); } } SortedBlocksWriter::TmpFilePtr SortedBlocksWriter::flush(const BlocksList & blocks) const { Pipes pipes; pipes.reserve(blocks.size()); for (const auto & block : blocks) if (auto num_rows = block.rows()) pipes.emplace_back(std::make_shared(block.cloneEmpty(), Chunk(block.getColumns(), num_rows))); if (pipes.empty()) return {}; QueryPipelineBuilder pipeline; pipeline.init(Pipe::unitePipes(std::move(pipes))); if (pipeline.getNumStreams() > 1) { auto transform = std::make_shared( pipeline.getHeader(), pipeline.getNumStreams(), sort_description, rows_in_block, SortingQueueStrategy::Default); pipeline.addTransform(std::move(transform)); } return flushToFile(volume->getDisk(), sample_block, std::move(pipeline), codec); } SortedBlocksWriter::PremergedFiles SortedBlocksWriter::premerge() { SortedFiles files; BlocksList blocks; /// wait other flushes if any { std::unique_lock lock{insert_mutex}; files.swap(sorted_files); blocks.swap(inserted_blocks.blocks); inserted_blocks.clear(); flush_condvar.wait(lock, [&]{ return !flush_inflight; }); } /// flush not flushed if (!blocks.empty()) files.emplace_back(flush(blocks)); Pipes pipes; pipes.reserve(num_files_for_merge); /// Merge by parts to save memory. It's possible to exchange disk I/O and memory by num_files_for_merge. { SortedFiles new_files; new_files.reserve(files.size() / num_files_for_merge + 1); while (files.size() > num_files_for_merge) { for (const auto & file : files) { pipes.emplace_back(streamFromFile(file)); if (pipes.size() == num_files_for_merge || &file == &files.back()) { QueryPipelineBuilder pipeline; pipeline.init(Pipe::unitePipes(std::move(pipes))); pipes = Pipes(); if (pipeline.getNumStreams() > 1) { auto transform = std::make_shared( pipeline.getHeader(), pipeline.getNumStreams(), sort_description, rows_in_block, SortingQueueStrategy::Default); pipeline.addTransform(std::move(transform)); } new_files.emplace_back(flushToFile(volume->getDisk(), sample_block, std::move(pipeline), codec)); } } files.clear(); files.swap(new_files); } for (const auto & file : files) pipes.emplace_back(streamFromFile(file)); } return PremergedFiles{std::move(files), Pipe::unitePipes(std::move(pipes))}; } SortedBlocksWriter::SortedFiles SortedBlocksWriter::finishMerge(std::function callback) { PremergedFiles files = premerge(); QueryPipelineBuilder pipeline; pipeline.init(std::move(files.pipe)); if (pipeline.getNumStreams() > 1) { ProfileEvents::increment(ProfileEvents::ExternalJoinMerge); auto transform = std::make_shared( pipeline.getHeader(), pipeline.getNumStreams(), sort_description, rows_in_block, SortingQueueStrategy::Default); pipeline.addTransform(std::move(transform)); } return flushToManyFiles(volume->getDisk(), sample_block, std::move(pipeline), codec, callback); } Pipe SortedBlocksWriter::streamFromFile(const TmpFilePtr & file) const { return Pipe(std::make_shared(file->path(), materializeBlock(sample_block))); } Block SortedBlocksBuffer::exchange(Block && block) { static constexpr const float reserve_coef = 1.2; Blocks out_blocks; Block empty_out = block.cloneEmpty(); { std::lock_guard lock(mutex); if (block) { current_bytes += block.bytes(); buffer.emplace_back(std::move(block)); /// Saved. Return empty block with same structure. if (current_bytes < max_bytes) return empty_out; } /// Not saved. Return buffered. out_blocks.swap(buffer); buffer.reserve(out_blocks.size() * reserve_coef); current_bytes = 0; } if (size_t size = out_blocks.size()) { if (size == 1) return out_blocks[0]; return mergeBlocks(std::move(out_blocks)); } return {}; } Block SortedBlocksBuffer::mergeBlocks(Blocks && blocks) const { size_t num_rows = 0; { /// Merge sort blocks Pipes pipes; pipes.reserve(blocks.size()); for (auto & block : blocks) { num_rows += block.rows(); Chunk chunk(block.getColumns(), block.rows()); pipes.emplace_back(std::make_shared(block.cloneEmpty(), std::move(chunk))); } Blocks tmp_blocks; QueryPipelineBuilder builder; builder.init(Pipe::unitePipes(std::move(pipes))); if (builder.getNumStreams() > 1) { auto transform = std::make_shared( builder.getHeader(), builder.getNumStreams(), sort_description, num_rows, SortingQueueStrategy::Default); builder.addTransform(std::move(transform)); } auto pipeline = QueryPipelineBuilder::getPipeline(std::move(builder)); PullingPipelineExecutor executor(pipeline); Block block; while (executor.pull(block)) tmp_blocks.emplace_back(block); blocks.swap(tmp_blocks); } if (blocks.size() == 1) return blocks[0]; return concatenateBlocks(blocks); } }