| 1 | #pragma once | 
|---|
| 2 |  | 
|---|
| 3 | #include <unordered_map> | 
|---|
| 4 |  | 
|---|
| 5 | #include <Core/Types.h> | 
|---|
| 6 |  | 
|---|
| 7 |  | 
|---|
| 8 | namespace DB | 
|---|
| 9 | { | 
|---|
| 10 |  | 
|---|
| 11 | class ReadBuffer; | 
|---|
| 12 | class WriteBuffer; | 
|---|
| 13 |  | 
|---|
| 14 | /** More information about the block. | 
|---|
| 15 | */ | 
|---|
| 16 | struct BlockInfo | 
|---|
| 17 | { | 
|---|
| 18 | /** is_overflows: | 
|---|
| 19 | * After running GROUP BY ... WITH TOTALS with the max_rows_to_group_by and group_by_overflow_mode = 'any' settings, | 
|---|
| 20 | *  a row is inserted in the separate block with aggregated values that have not passed max_rows_to_group_by. | 
|---|
| 21 | * If it is such a block, then is_overflows is set to true for it. | 
|---|
| 22 | */ | 
|---|
| 23 |  | 
|---|
| 24 | /** bucket_num: | 
|---|
| 25 | * When using the two-level aggregation method, data with different key groups are scattered across different buckets. | 
|---|
| 26 | * In this case, the bucket number is indicated here. It is used to optimize the merge for distributed aggregation. | 
|---|
| 27 | * Otherwise -1. | 
|---|
| 28 | */ | 
|---|
| 29 |  | 
|---|
| 30 | #define APPLY_FOR_BLOCK_INFO_FIELDS(M) \ | 
|---|
| 31 | M(bool,     is_overflows,     false,     1) \ | 
|---|
| 32 | M(Int32,    bucket_num,     -1,     2) | 
|---|
| 33 |  | 
|---|
| 34 | #define DECLARE_FIELD(TYPE, NAME, DEFAULT, FIELD_NUM) \ | 
|---|
| 35 | TYPE NAME = DEFAULT; | 
|---|
| 36 |  | 
|---|
| 37 | APPLY_FOR_BLOCK_INFO_FIELDS(DECLARE_FIELD) | 
|---|
| 38 |  | 
|---|
| 39 | #undef DECLARE_FIELD | 
|---|
| 40 |  | 
|---|
| 41 | /// Write the values in binary form. NOTE: You could use protobuf, but it would be overkill for this case. | 
|---|
| 42 | void write(WriteBuffer & out) const; | 
|---|
| 43 |  | 
|---|
| 44 | /// Read the values in binary form. | 
|---|
| 45 | void read(ReadBuffer & in); | 
|---|
| 46 | }; | 
|---|
| 47 |  | 
|---|
| 48 | /// Block extention to support delayed defaults. AddingDefaultsBlockInputStream uses it to replace missing values with column defaults. | 
|---|
| 49 | class BlockMissingValues | 
|---|
| 50 | { | 
|---|
| 51 | public: | 
|---|
| 52 | using RowsBitMask = std::vector<bool>; /// a bit per row for a column | 
|---|
| 53 |  | 
|---|
| 54 | const RowsBitMask & getDefaultsBitmask(size_t column_idx) const; | 
|---|
| 55 | void setBit(size_t column_idx, size_t row_idx); | 
|---|
| 56 | bool empty() const { return rows_mask_by_column_id.empty(); } | 
|---|
| 57 | size_t size() const { return rows_mask_by_column_id.size(); } | 
|---|
| 58 | void clear() { rows_mask_by_column_id.clear(); } | 
|---|
| 59 |  | 
|---|
| 60 | private: | 
|---|
| 61 | using RowsMaskByColumnId = std::unordered_map<size_t, RowsBitMask>; | 
|---|
| 62 |  | 
|---|
| 63 | /// If rows_mask_by_column_id[column_id][row_id] is true related value in Block should be replaced with column default. | 
|---|
| 64 | /// It could contain less columns and rows then related block. | 
|---|
| 65 | RowsMaskByColumnId rows_mask_by_column_id; | 
|---|
| 66 | }; | 
|---|
| 67 |  | 
|---|
| 68 | } | 
|---|
| 69 |  | 
|---|