ClickHouse/src/Storages/MergeTree/MergeTask.h

#pragma once

#include <Storages/MergeTree/IExecutableTask.h>
#include <Storages/MergeTree/MergeProgress.h>
#include <Storages/MergeTree/MergeTreeData.h>
#include <Storages/MergeTree/IMergedBlockOutputStream.h>
#include <Storages/MergeTree/MergedBlockOutputStream.h>
#include <Storages/MergeTree/FutureMergedMutatedPart.h>
#include <Storages/MergeTree/ColumnSizeEstimator.h>
#include <Storages/MergeTree/MergedColumnOnlyOutputStream.h>
#include <Processors/Transforms/ColumnGathererTransform.h>
#include <Processors/Executors/PullingPipelineExecutor.h>
#include <QueryPipeline/QueryPipeline.h>
#include <Compression/CompressedReadBufferFromFile.h>
#include <Common/filesystemHelpers.h>

#include <memory>
#include <list>

namespace DB
{

class MergeTask;
using MergeTaskPtr = std::shared_ptr<MergeTask>;

/**
 * Overview of the merge algorithm
 *
 * Each merge is executed sequentially block by block.
 * The main idea is to make a merge not a subroutine which is executed
 * in a thread pool and may occupy a thread for a period of time,
 * but to make a merge a coroutine which can suspend the execution
 * in some points and then resume the execution from this point.
 *
 * A perfect point where to suspend the execution is after the work over a block is finished.
 * The task itself will be executed via BackgroundJobExecutor.
 *
 * The interface of the task is simple.
 * The main method is `execute()` which will return true, if the task wants to be executed again and false otherwise.
 *
 * With this kind of task we can give a merge a priority.
 * A priority is simple - the lower the size of the merge, the higher priority.
 * So, if ClickHouse wants to merge some really big parts into a bigger part,
 * then it will be executed for a long time, because the result of the merge is not really needed immediately.
 * It is better to merge small parts as soon as possible.
*/
class MergeTask
{
public:

    MergeTask(
        FutureMergedMutatedPartPtr future_part_,
        StorageMetadataPtr metadata_snapshot_,
        MergeList::Entry * merge_entry_,
        std::unique_ptr<MergeListElement> projection_merge_list_element_,
        time_t time_of_merge_,
        ContextPtr context_,
        ReservationSharedPtr space_reservation_,
        bool deduplicate_,
        Names deduplicate_by_columns_,
        MergeTreeData::MergingParams merging_params_,
        bool need_prefix,
        IMergeTreeDataPart * parent_part_,
        String suffix_,
        MergeTreeTransactionPtr txn,
        MergeTreeData * data_,
        MergeTreeDataMergerMutator * mutator_,
        ActionBlocker * merges_blocker_,
        ActionBlocker * ttl_merges_blocker_)
        {
            global_ctx = std::make_shared<GlobalRuntimeContext>();

            global_ctx->future_part = std::move(future_part_);
            global_ctx->metadata_snapshot = std::move(metadata_snapshot_);
            global_ctx->merge_entry = std::move(merge_entry_);
            global_ctx->projection_merge_list_element = std::move(projection_merge_list_element_);
            global_ctx->merge_list_element_ptr
                = global_ctx->projection_merge_list_element ? global_ctx->projection_merge_list_element.get() : (*global_ctx->merge_entry)->ptr();
            global_ctx->time_of_merge = std::move(time_of_merge_);
            global_ctx->context = std::move(context_);
            global_ctx->space_reservation = std::move(space_reservation_);
            global_ctx->deduplicate = std::move(deduplicate_);
            global_ctx->deduplicate_by_columns = std::move(deduplicate_by_columns_);
            global_ctx->parent_part = std::move(parent_part_);
            global_ctx->data = std::move(data_);
            global_ctx->mutator = std::move(mutator_);
            global_ctx->merges_blocker = std::move(merges_blocker_);
            global_ctx->ttl_merges_blocker = std::move(ttl_merges_blocker_);
            global_ctx->txn = std::move(txn);
            global_ctx->need_prefix = need_prefix;

            auto prepare_stage_ctx = std::make_shared<ExecuteAndFinalizeHorizontalPartRuntimeContext>();

            prepare_stage_ctx->suffix = std::move(suffix_);
            prepare_stage_ctx->merging_params = std::move(merging_params_);

            (*stages.begin())->setRuntimeContext(std::move(prepare_stage_ctx), global_ctx);
        }

    std::future<MergeTreeData::MutableDataPartPtr> getFuture()
    {
        return global_ctx->promise.get_future();
    }

    bool execute();

private:
    struct IStage;
    using StagePtr = std::shared_ptr<IStage>;

    struct IStageRuntimeContext {};
    using StageRuntimeContextPtr = std::shared_ptr<IStageRuntimeContext>;

    struct IStage
    {
        virtual void setRuntimeContext(StageRuntimeContextPtr local, StageRuntimeContextPtr global) = 0;
        virtual StageRuntimeContextPtr getContextForNextStage() = 0;
        virtual bool execute() = 0;
        virtual ~IStage() = default;
    };

    /// By default this context is uninitialed, but some variables has to be set after construction,
    /// some variables are used in a process of execution
    /// Proper initialization is responsibility of the author
    struct GlobalRuntimeContext : public IStageRuntimeContext //-V730
    {
        MergeList::Entry * merge_entry{nullptr};
        /// If not null, use this instead of the global MergeList::Entry. This is for merging projections.
        std::unique_ptr<MergeListElement> projection_merge_list_element;
        MergeListElement * merge_list_element_ptr{nullptr};
        MergeTreeData * data{nullptr};
        MergeTreeDataMergerMutator * mutator{nullptr};
        ActionBlocker * merges_blocker{nullptr};
        ActionBlocker * ttl_merges_blocker{nullptr};
        StorageSnapshotPtr storage_snapshot{nullptr};
        StorageMetadataPtr metadata_snapshot{nullptr};
        FutureMergedMutatedPartPtr future_part{nullptr};
        /// This will be either nullptr or new_data_part, so raw pointer is ok.
        IMergeTreeDataPart * parent_part{nullptr};
        ContextPtr context{nullptr};
        time_t time_of_merge{0};
        ReservationSharedPtr space_reservation{nullptr};
        bool deduplicate{false};
        Names deduplicate_by_columns{};

        NamesAndTypesList gathering_columns{};
        NamesAndTypesList merging_columns{};
        Names gathering_column_names{};
        Names merging_column_names{};
        NamesAndTypesList storage_columns{};
        Names all_column_names{};
        MergeTreeData::DataPart::Checksums checksums_gathered_columns{};

        MergeAlgorithm chosen_merge_algorithm{MergeAlgorithm::Undecided};
        size_t gathering_column_names_size{0};

        std::unique_ptr<MergeStageProgress> horizontal_stage_progress{nullptr};
        std::unique_ptr<MergeStageProgress> column_progress{nullptr};

        std::shared_ptr<MergedBlockOutputStream> to{nullptr};
        QueryPipeline merged_pipeline;
        std::unique_ptr<PullingPipelineExecutor> merging_executor;

        MergeTreeData::MutableDataPartPtr new_data_part{nullptr};

        /// If lightweight delete mask is present then some input rows are filtered out right after reading.
        std::shared_ptr<std::atomic<size_t>> input_rows_filtered{std::make_shared<std::atomic<size_t>>(0)};
        size_t rows_written{0};
        UInt64 watch_prev_elapsed{0};

        std::promise<MergeTreeData::MutableDataPartPtr> promise{};

        IMergedBlockOutputStream::WrittenOffsetColumns written_offset_columns{};

        MergeTreeTransactionPtr txn;
        bool need_prefix;

        scope_guard temporary_directory_lock;
    };

    using GlobalRuntimeContextPtr = std::shared_ptr<GlobalRuntimeContext>;

    /// By default this context is uninitialed, but some variables has to be set after construction,
    /// some variables are used in a process of execution
    /// Proper initialization is responsibility of the author
    struct ExecuteAndFinalizeHorizontalPartRuntimeContext : public IStageRuntimeContext //-V730
    {
        /// Dependencies
        String suffix;
        bool need_prefix;
        MergeTreeData::MergingParams merging_params{};

        DiskPtr tmp_disk{nullptr};
        DiskPtr disk{nullptr};
        bool need_remove_expired_values{false};
        bool force_ttl{false};
        CompressionCodecPtr compression_codec{nullptr};
        size_t sum_input_rows_upper_bound{0};
        std::unique_ptr<PocoTemporaryFile> rows_sources_file{nullptr};
        std::unique_ptr<WriteBufferFromFileBase> rows_sources_uncompressed_write_buf{nullptr};
        std::unique_ptr<WriteBuffer> rows_sources_write_buf{nullptr};
        std::optional<ColumnSizeEstimator> column_sizes{};

        size_t initial_reservation{0};
        bool read_with_direct_io{false};

        std::function<bool()> is_cancelled{};

        /// Local variables for this stage
        size_t sum_compressed_bytes_upper_bound{0};
        bool blocks_are_granules_size{false};

        Poco::Logger * log{&Poco::Logger::get("MergeTask::PrepareStage")};

        /// Dependencies for next stages
        std::list<DB::NameAndTypePair>::const_iterator it_name_and_type;
        size_t column_num_for_vertical_merge{0};
        bool need_sync{false};
    };

    using ExecuteAndFinalizeHorizontalPartRuntimeContextPtr = std::shared_ptr<ExecuteAndFinalizeHorizontalPartRuntimeContext>;


    struct ExecuteAndFinalizeHorizontalPart : public IStage
    {
        bool execute() override;

        bool prepare();
        bool executeImpl();

        using ExecuteAndFinalizeHorizontalPartSubtasks = std::array<std::function<bool()>, 2>;

        ExecuteAndFinalizeHorizontalPartSubtasks subtasks
        {
            [this] () { return prepare(); },
            [this] () { return executeImpl(); }
        };

        ExecuteAndFinalizeHorizontalPartSubtasks::iterator subtasks_iterator = subtasks.begin();


        MergeAlgorithm chooseMergeAlgorithm() const;
        void createMergedStream();

        void setRuntimeContext(StageRuntimeContextPtr local, StageRuntimeContextPtr global) override
        {
            ctx = static_pointer_cast<ExecuteAndFinalizeHorizontalPartRuntimeContext>(local);
            global_ctx = static_pointer_cast<GlobalRuntimeContext>(global);
        }
        StageRuntimeContextPtr getContextForNextStage() override;

        ExecuteAndFinalizeHorizontalPartRuntimeContextPtr ctx;
        GlobalRuntimeContextPtr global_ctx;
    };

    /// By default this context is uninitialed, but some variables has to be set after construction,
    /// some variables are used in a process of execution
    /// Proper initialization is responsibility of the author
    struct VerticalMergeRuntimeContext : public IStageRuntimeContext //-V730
    {
        /// Begin dependencies from previous stage
        std::unique_ptr<PocoTemporaryFile> rows_sources_file;
        std::unique_ptr<WriteBufferFromFileBase> rows_sources_uncompressed_write_buf{nullptr};
        std::unique_ptr<WriteBuffer> rows_sources_write_buf{nullptr};
        std::optional<ColumnSizeEstimator> column_sizes;
        CompressionCodecPtr compression_codec;
        DiskPtr tmp_disk{nullptr};
        std::list<DB::NameAndTypePair>::const_iterator it_name_and_type;
        size_t column_num_for_vertical_merge{0};
        bool read_with_direct_io{false};
        bool need_sync{false};
        /// End dependencies from previous stages

        enum class State
        {
            NEED_PREPARE,
            NEED_EXECUTE,
            NEED_FINISH
        };
        State vertical_merge_one_column_state{State::NEED_PREPARE};

        Float64 progress_before = 0;
        std::unique_ptr<MergedColumnOnlyOutputStream> column_to{nullptr};
        std::vector<std::unique_ptr<MergedColumnOnlyOutputStream>> delayed_streams;
        size_t column_elems_written{0};
        QueryPipeline column_parts_pipeline;
        std::unique_ptr<PullingPipelineExecutor> executor;
        std::unique_ptr<CompressedReadBufferFromFile> rows_sources_read_buf{nullptr};
    };

    using VerticalMergeRuntimeContextPtr = std::shared_ptr<VerticalMergeRuntimeContext>;


    struct VerticalMergeStage : public IStage
    {
        bool execute() override;
        void setRuntimeContext(StageRuntimeContextPtr local, StageRuntimeContextPtr global) override
        {
            ctx = static_pointer_cast<VerticalMergeRuntimeContext>(local);
            global_ctx = static_pointer_cast<GlobalRuntimeContext>(global);
        }
        StageRuntimeContextPtr getContextForNextStage() override;

        bool prepareVerticalMergeForAllColumns() const;
        bool executeVerticalMergeForAllColumns() const;
        bool finalizeVerticalMergeForAllColumns() const;

        using VerticalMergeStageSubtasks = std::array<std::function<bool()>, 3>;

        VerticalMergeStageSubtasks subtasks
        {
            [this] () { return prepareVerticalMergeForAllColumns(); },
            [this] () { return executeVerticalMergeForAllColumns(); },
            [this] () { return finalizeVerticalMergeForAllColumns(); }
        };

        VerticalMergeStageSubtasks::iterator subtasks_iterator = subtasks.begin();

        void prepareVerticalMergeForOneColumn() const;
        bool executeVerticalMergeForOneColumn() const;
        void finalizeVerticalMergeForOneColumn() const;

        VerticalMergeRuntimeContextPtr ctx;
        GlobalRuntimeContextPtr global_ctx;
    };

    /// By default this context is uninitialed, but some variables has to be set after construction,
    /// some variables are used in a process of execution
    /// Proper initialization is responsibility of the author
    struct MergeProjectionsRuntimeContext : public IStageRuntimeContext //-V730
    {
        /// Only one dependency
        bool need_sync{false};

        using MergeTasks = std::deque<MergeTaskPtr>;
        MergeTasks tasks_for_projections;
        MergeTasks::iterator projections_iterator;

        Poco::Logger * log{&Poco::Logger::get("MergeTask::MergeProjectionsStage")};
    };

    using MergeProjectionsRuntimeContextPtr = std::shared_ptr<MergeProjectionsRuntimeContext>;

    struct MergeProjectionsStage : public IStage
    {
        bool execute() override;
        void setRuntimeContext(StageRuntimeContextPtr local, StageRuntimeContextPtr global) override
        {
            ctx = static_pointer_cast<MergeProjectionsRuntimeContext>(local);
            global_ctx = static_pointer_cast<GlobalRuntimeContext>(global);
        }
        StageRuntimeContextPtr getContextForNextStage() override { return nullptr; }

        bool mergeMinMaxIndexAndPrepareProjections() const;
        bool executeProjections() const;
        bool finalizeProjectionsAndWholeMerge() const;

        using MergeProjectionsStageSubtasks = std::array<std::function<bool()>, 3>;

        MergeProjectionsStageSubtasks subtasks
        {
            [this] () { return mergeMinMaxIndexAndPrepareProjections(); },
            [this] () { return executeProjections(); },
            [this] () { return finalizeProjectionsAndWholeMerge(); }
        };

        MergeProjectionsStageSubtasks::iterator subtasks_iterator = subtasks.begin();

        MergeProjectionsRuntimeContextPtr ctx;
        GlobalRuntimeContextPtr global_ctx;
    };

    GlobalRuntimeContextPtr global_ctx;

    using Stages = std::array<StagePtr, 3>;

    Stages stages
    {
        std::make_shared<ExecuteAndFinalizeHorizontalPart>(),
        std::make_shared<VerticalMergeStage>(),
        std::make_shared<MergeProjectionsStage>()
    };

    Stages::iterator stages_iterator = stages.begin();

};

/// FIXME
[[ maybe_unused]] static MergeTreeData::MutableDataPartPtr executeHere(MergeTaskPtr task)
{
    while (task->execute()) {}
    return task->getFuture().get();
}

}