#pragma once #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include namespace DB { namespace ErrorCodes { extern const int ILLEGAL_COLUMN; extern const int BAD_ARGUMENTS; } struct BloomFilterHash { static constexpr UInt64 bf_hash_seed[15] = { 13635471485423070496ULL, 10336109063487487899ULL, 17779957404565211594ULL, 8988612159822229247ULL, 4954614162757618085ULL, 12980113590177089081ULL, 9263883436177860930ULL, 3656772712723269762ULL, 10362091744962961274ULL, 7582936617938287249ULL, 15033938188484401405ULL, 18286745649494826751ULL, 6852245486148412312ULL, 8886056245089344681ULL, 10151472371158292780ULL }; template static UInt64 getNumberTypeHash(const Field & field) { /// For negative, we should convert the type to make sure the symbol is in right place return field.isNull() ? intHash64(0) : intHash64(ext::bit_cast(FieldType(field.safeGet()))); } static UInt64 getStringTypeHash(const Field & field) { if (!field.isNull()) { const auto & value = field.safeGet(); return CityHash_v1_0_2::CityHash64(value.data(), value.size()); } return CityHash_v1_0_2::CityHash64("", 0); } static UInt64 getFixedStringTypeHash(const Field & field, const IDataType * type) { if (!field.isNull()) { const auto & value = field.safeGet(); return CityHash_v1_0_2::CityHash64(value.data(), value.size()); } const auto * fixed_string_type = typeid_cast(type); const std::vector value(fixed_string_type->getN(), 0); return CityHash_v1_0_2::CityHash64(value.data(), value.size()); } static ColumnPtr hashWithField(const IDataType * data_type, const Field & field) { const auto & build_hash_column = [&](const UInt64 & hash) -> ColumnPtr { return ColumnConst::create(ColumnUInt64::create(1, hash), 1); }; WhichDataType which(data_type); if (which.isUInt8()) return build_hash_column(getNumberTypeHash(field)); else if (which.isUInt16()) return build_hash_column(getNumberTypeHash(field)); else if (which.isUInt32()) return build_hash_column(getNumberTypeHash(field)); else if (which.isUInt64()) return build_hash_column(getNumberTypeHash(field)); else if (which.isInt8()) return build_hash_column(getNumberTypeHash(field)); else if (which.isInt16()) return build_hash_column(getNumberTypeHash(field)); else if (which.isInt32()) return build_hash_column(getNumberTypeHash(field)); else if (which.isInt64()) return build_hash_column(getNumberTypeHash(field)); else if (which.isEnum8()) return build_hash_column(getNumberTypeHash(field)); else if (which.isEnum16()) return build_hash_column(getNumberTypeHash(field)); else if (which.isDate()) return build_hash_column(getNumberTypeHash(field)); else if (which.isDateTime()) return build_hash_column(getNumberTypeHash(field)); else if (which.isFloat32()) return build_hash_column(getNumberTypeHash(field)); else if (which.isFloat64()) return build_hash_column(getNumberTypeHash(field)); else if (which.isString()) return build_hash_column(getStringTypeHash(field)); else if (which.isFixedString()) return build_hash_column(getFixedStringTypeHash(field, data_type)); else throw Exception("Unexpected type " + data_type->getName() + " of bloom filter index.", ErrorCodes::BAD_ARGUMENTS); } static ColumnPtr hashWithColumn(const DataTypePtr & data_type, const ColumnPtr & column, size_t pos, size_t limit) { WhichDataType which(data_type); if (which.isArray()) { const auto * array_col = typeid_cast(column.get()); if (checkAndGetColumn(array_col->getData())) throw Exception("Unexpected type " + data_type->getName() + " of bloom filter index.", ErrorCodes::BAD_ARGUMENTS); const auto & offsets = array_col->getOffsets(); limit = offsets[pos + limit - 1] - offsets[pos - 1]; /// PaddedPODArray allows access on index -1. pos = offsets[pos - 1]; if (limit == 0) { auto index_column = ColumnUInt64::create(1); ColumnUInt64::Container & index_column_vec = index_column->getData(); index_column_vec[0] = 0; return index_column; } } const ColumnPtr actual_col = BloomFilter::getPrimitiveColumn(column); const DataTypePtr actual_type = BloomFilter::getPrimitiveType(data_type); auto index_column = ColumnUInt64::create(limit); ColumnUInt64::Container & index_column_vec = index_column->getData(); getAnyTypeHash(actual_type.get(), actual_col.get(), index_column_vec, pos); return index_column; } template static void getAnyTypeHash(const IDataType * data_type, const IColumn * column, ColumnUInt64::Container & vec, size_t pos) { WhichDataType which(data_type); if (which.isUInt8()) getNumberTypeHash(column, vec, pos); else if (which.isUInt16()) getNumberTypeHash(column, vec, pos); else if (which.isUInt32()) getNumberTypeHash(column, vec, pos); else if (which.isUInt64()) getNumberTypeHash(column, vec, pos); else if (which.isInt8()) getNumberTypeHash(column, vec, pos); else if (which.isInt16()) getNumberTypeHash(column, vec, pos); else if (which.isInt32()) getNumberTypeHash(column, vec, pos); else if (which.isInt64()) getNumberTypeHash(column, vec, pos); else if (which.isEnum8()) getNumberTypeHash(column, vec, pos); else if (which.isEnum16()) getNumberTypeHash(column, vec, pos); else if (which.isDate()) getNumberTypeHash(column, vec, pos); else if (which.isDateTime()) getNumberTypeHash(column, vec, pos); else if (which.isFloat32()) getNumberTypeHash(column, vec, pos); else if (which.isFloat64()) getNumberTypeHash(column, vec, pos); else if (which.isString()) getStringTypeHash(column, vec, pos); else if (which.isFixedString()) getStringTypeHash(column, vec, pos); else throw Exception("Unexpected type " + data_type->getName() + " of bloom filter index.", ErrorCodes::BAD_ARGUMENTS); } template static void getNumberTypeHash(const IColumn * column, ColumnUInt64::Container & vec, size_t pos) { const auto * index_column = typeid_cast *>(column); if (unlikely(!index_column)) throw Exception("Illegal column type was passed to the bloom filter index.", ErrorCodes::ILLEGAL_COLUMN); const typename ColumnVector::Container & vec_from = index_column->getData(); /// Because we're missing the precision of float in the Field.h /// to be consistent, we need to convert Float32 to Float64 processing, also see: BloomFilterHash::hashWithField if constexpr (std::is_same_v, ColumnFloat32>) { for (size_t index = 0, size = vec.size(); index < size; ++index) { UInt64 hash = intHash64(ext::bit_cast(Float64(vec_from[index + pos]))); if constexpr (is_first) vec[index] = hash; else vec[index] = CityHash_v1_0_2::Hash128to64(CityHash_v1_0_2::uint128(vec[index], hash)); } } else { for (size_t index = 0, size = vec.size(); index < size; ++index) { UInt64 hash = intHash64(ext::bit_cast(vec_from[index + pos])); if constexpr (is_first) vec[index] = hash; else vec[index] = CityHash_v1_0_2::Hash128to64(CityHash_v1_0_2::uint128(vec[index], hash)); } } } template static void getStringTypeHash(const IColumn * column, ColumnUInt64::Container & vec, size_t pos) { if (const auto * index_column = typeid_cast(column)) { const ColumnString::Chars & data = index_column->getChars(); const ColumnString::Offsets & offsets = index_column->getOffsets(); ColumnString::Offset current_offset = pos; for (size_t index = 0, size = vec.size(); index < size; ++index) { UInt64 city_hash = CityHash_v1_0_2::CityHash64( reinterpret_cast(&data[current_offset]), offsets[index + pos] - current_offset - 1); if constexpr (is_first) vec[index] = city_hash; else vec[index] = CityHash_v1_0_2::Hash128to64(CityHash_v1_0_2::uint128(vec[index], city_hash)); current_offset = offsets[index + pos]; } } else if (const auto * fixed_string_index_column = typeid_cast(column)) { size_t fixed_len = fixed_string_index_column->getN(); const auto & data = fixed_string_index_column->getChars(); for (size_t index = 0, size = vec.size(); index < size; ++index) { UInt64 city_hash = CityHash_v1_0_2::CityHash64(reinterpret_cast(&data[(index + pos) * fixed_len]), fixed_len); if constexpr (is_first) vec[index] = city_hash; else vec[index] = CityHash_v1_0_2::Hash128to64(CityHash_v1_0_2::uint128(vec[index], city_hash)); } } else throw Exception("Illegal column type was passed to the bloom filter index.", ErrorCodes::ILLEGAL_COLUMN); } static std::pair calculationBestPractices(double max_conflict_probability) { static const size_t MAX_BITS_PER_ROW = 20; static const size_t MAX_HASH_FUNCTION_COUNT = 15; /// For the smallest index per level in probability_lookup_table static const size_t min_probability_index_each_bits[] = {0, 0, 1, 2, 3, 3, 4, 5, 6, 6, 7, 8, 8, 9, 10, 10, 11, 12, 12, 13, 14}; static const long double probability_lookup_table[MAX_BITS_PER_ROW + 1][MAX_HASH_FUNCTION_COUNT] = { {1.0}, /// dummy, 0 bits per row {1.0, 1.0}, {1.0, 0.393, 0.400}, {1.0, 0.283, 0.237, 0.253}, {1.0, 0.221, 0.155, 0.147, 0.160}, {1.0, 0.181, 0.109, 0.092, 0.092, 0.101}, // 5 {1.0, 0.154, 0.0804, 0.0609, 0.0561, 0.0578, 0.0638}, {1.0, 0.133, 0.0618, 0.0423, 0.0359, 0.0347, 0.0364}, {1.0, 0.118, 0.0489, 0.0306, 0.024, 0.0217, 0.0216, 0.0229}, {1.0, 0.105, 0.0397, 0.0228, 0.0166, 0.0141, 0.0133, 0.0135, 0.0145}, {1.0, 0.0952, 0.0329, 0.0174, 0.0118, 0.00943, 0.00844, 0.00819, 0.00846}, // 10 {1.0, 0.0869, 0.0276, 0.0136, 0.00864, 0.0065, 0.00552, 0.00513, 0.00509}, {1.0, 0.08, 0.0236, 0.0108, 0.00646, 0.00459, 0.00371, 0.00329, 0.00314}, {1.0, 0.074, 0.0203, 0.00875, 0.00492, 0.00332, 0.00255, 0.00217, 0.00199, 0.00194}, {1.0, 0.0689, 0.0177, 0.00718, 0.00381, 0.00244, 0.00179, 0.00146, 0.00129, 0.00121, 0.0012}, {1.0, 0.0645, 0.0156, 0.00596, 0.003, 0.00183, 0.00128, 0.001, 0.000852, 0.000775, 0.000744}, // 15 {1.0, 0.0606, 0.0138, 0.005, 0.00239, 0.00139, 0.000935, 0.000702, 0.000574, 0.000505, 0.00047, 0.000459}, {1.0, 0.0571, 0.0123, 0.00423, 0.00193, 0.00107, 0.000692, 0.000499, 0.000394, 0.000335, 0.000302, 0.000287, 0.000284}, {1.0, 0.054, 0.0111, 0.00362, 0.00158, 0.000839, 0.000519, 0.00036, 0.000275, 0.000226, 0.000198, 0.000183, 0.000176}, {1.0, 0.0513, 0.00998, 0.00312, 0.0013, 0.000663, 0.000394, 0.000264, 0.000194, 0.000155, 0.000132, 0.000118, 0.000111, 0.000109}, {1.0, 0.0488, 0.00906, 0.0027, 0.00108, 0.00053, 0.000303, 0.000196, 0.00014, 0.000108, 8.89e-05, 7.77e-05, 7.12e-05, 6.79e-05, 6.71e-05} // 20 }; for (size_t bits_per_row = 1; bits_per_row < MAX_BITS_PER_ROW; ++bits_per_row) { if (probability_lookup_table[bits_per_row][min_probability_index_each_bits[bits_per_row]] <= max_conflict_probability) { size_t max_size_of_hash_functions = min_probability_index_each_bits[bits_per_row]; for (size_t size_of_hash_functions = max_size_of_hash_functions; size_of_hash_functions > 0; --size_of_hash_functions) if (probability_lookup_table[bits_per_row][size_of_hash_functions] > max_conflict_probability) return std::pair(bits_per_row, size_of_hash_functions + 1); } } return std::pair(MAX_BITS_PER_ROW - 1, min_probability_index_each_bits[MAX_BITS_PER_ROW - 1]); } }; }