dbms: Server: Feature implementation. [#METR-16188]

This commit is contained in:
Alexey Arno 2015-05-18 01:33:17 +03:00
parent bf6aecc826
commit 5f0a1cab74
2 changed files with 77 additions and 17 deletions

View File

@ -9,13 +9,9 @@
#include <cmath> #include <cmath>
struct C;
struct C;
namespace DB namespace DB
{ {
/// XXX Реализовать корреляцию (corr).
/** Статистические аггрегатные функции: /** Статистические аггрегатные функции:
* varSamp - выборочная дисперсия * varSamp - выборочная дисперсия
* stddevSamp - среднее выборочное квадратичное отклонение * stddevSamp - среднее выборочное квадратичное отклонение
@ -23,6 +19,7 @@ namespace DB
* stddevPop - среднее квадратичное отклонение * stddevPop - среднее квадратичное отклонение
* covarSamp - выборочная ковариация * covarSamp - выборочная ковариация
* covarPop - ковариация * covarPop - ковариация
* corr - корреляция
*/ */
/** Параллельный и инкрементальный алгоритм для вычисления дисперсии. /** Параллельный и инкрементальный алгоритм для вычисления дисперсии.
@ -62,7 +59,7 @@ public:
mean = (source.count * source.mean + count * mean) / total_count; mean = (source.count * source.mean + count * mean) / total_count;
} }
else else
mean = source.mean + delta * (count / total_count); mean = source.mean + delta * (static_cast<Float64>(count) / total_count);
m2 += source.m2 + delta * delta * factor; m2 += source.m2 + delta * delta * factor;
count = total_count; count = total_count;
@ -206,7 +203,7 @@ struct StdDevPopImpl
* (J. Bennett et al., Sandia National Laboratories, * (J. Bennett et al., Sandia National Laboratories,
* 2009 IEEE International Conference on Cluster Computing) * 2009 IEEE International Conference on Cluster Computing)
*/ */
template<typename T, typename U, typename Op> template<typename T, typename U, typename Op, bool compute_marginal_moments>
class CovarianceData class CovarianceData
{ {
public: public:
@ -216,16 +213,25 @@ public:
{ {
T left_received = static_cast<const ColumnVector<T> &>(column_left).getData()[row_num]; T left_received = static_cast<const ColumnVector<T> &>(column_left).getData()[row_num];
Float64 val_left = static_cast<Float64>(left_received); Float64 val_left = static_cast<Float64>(left_received);
Float64 left_delta = val_left - left_mean;
U right_received = static_cast<const ColumnVector<U> &>(column_right).getData()[row_num]; U right_received = static_cast<const ColumnVector<U> &>(column_right).getData()[row_num];
Float64 val_right = static_cast<Float64>(right_received); Float64 val_right = static_cast<Float64>(right_received);
Float64 right_delta = val_right - right_mean;
Float64 old_right_mean = right_mean; Float64 old_right_mean = right_mean;
++count; ++count;
left_mean += (val_left - left_mean) / count;
right_mean += (val_right - right_mean) / count; left_mean += left_delta / count;
right_mean += right_delta / count;
co_moment += (val_left - left_mean) * (val_right - old_right_mean); co_moment += (val_left - left_mean) * (val_right - old_right_mean);
if (compute_marginal_moments)
{
left_m2 += left_delta * (val_left - left_mean);
right_m2 += right_delta * (val_right - right_mean);
}
} }
void mergeWith(const CovarianceData & source) void mergeWith(const CovarianceData & source)
@ -238,10 +244,16 @@ public:
Float64 left_delta = left_mean - source.left_mean; Float64 left_delta = left_mean - source.left_mean;
Float64 right_delta = right_mean - source.right_mean; Float64 right_delta = right_mean - source.right_mean;
left_mean += left_delta * (source.count / total_count); left_mean += left_delta * (static_cast<Float64>(source.count) / total_count);
right_mean += right_delta * (source.count / total_count); right_mean += right_delta * (static_cast<Float64>(source.count) / total_count);
co_moment += source.co_moment + left_delta * right_delta * factor; co_moment += source.co_moment + left_delta * right_delta * factor;
count = total_count; count = total_count;
if (compute_marginal_moments)
{
left_m2 += source.left_m2 + left_delta * left_delta * factor;
right_m2 += source.right_m2 + right_delta * right_delta * factor;
}
} }
void serialize(WriteBuffer & buf) const void serialize(WriteBuffer & buf) const
@ -250,6 +262,12 @@ public:
writeBinary(left_mean, buf); writeBinary(left_mean, buf);
writeBinary(right_mean, buf); writeBinary(right_mean, buf);
writeBinary(co_moment, buf); writeBinary(co_moment, buf);
if (compute_marginal_moments)
{
writeBinary(left_m2, buf);
writeBinary(right_m2, buf);
}
} }
void deserialize(ReadBuffer & buf) void deserialize(ReadBuffer & buf)
@ -258,11 +276,17 @@ public:
readBinary(left_mean, buf); readBinary(left_mean, buf);
readBinary(right_mean, buf); readBinary(right_mean, buf);
readBinary(co_moment, buf); readBinary(co_moment, buf);
if (compute_marginal_moments)
{
readBinary(left_m2, buf);
readBinary(right_m2, buf);
}
} }
void publish(IColumn & to) const void publish(IColumn & to) const
{ {
static_cast<ColumnFloat64 &>(to).getData().push_back(Op::apply(co_moment, count)); static_cast<ColumnFloat64 &>(to).getData().push_back(Op::apply(co_moment, left_m2, right_m2, count));
} }
private: private:
@ -270,10 +294,15 @@ private:
Float64 left_mean = 0.0; Float64 left_mean = 0.0;
Float64 right_mean = 0.0; Float64 right_mean = 0.0;
Float64 co_moment = 0.0; Float64 co_moment = 0.0;
Float64 left_m2 = 0.0;
Float64 right_m2 = 0.0;
}; };
template<typename T, typename U, typename Op> template<typename T, typename U, typename Op, bool compute_marginal_moments = false>
class AggregateFunctionCovariance final : public IBinaryAggregateFunction<CovarianceData<T, U, Op>, AggregateFunctionCovariance<T, U, Op> > class AggregateFunctionCovariance final
: public IBinaryAggregateFunction<
CovarianceData<T, U, Op, compute_marginal_moments>,
AggregateFunctionCovariance<T, U, Op, compute_marginal_moments> >
{ {
public: public:
String getName() const override { return Op::name; } String getName() const override { return Op::name; }
@ -311,7 +340,7 @@ public:
void deserializeMerge(AggregateDataPtr place, ReadBuffer & buf) const override void deserializeMerge(AggregateDataPtr place, ReadBuffer & buf) const override
{ {
CovarianceData<T, U, Op> source; CovarianceData<T, U, Op, compute_marginal_moments> source;
source.deserialize(buf); source.deserialize(buf);
this->data(place).mergeWith(source); this->data(place).mergeWith(source);
@ -332,7 +361,7 @@ struct CovarSampImpl
{ {
static constexpr auto name = "covarSamp"; static constexpr auto name = "covarSamp";
static inline Float64 apply(Float64 co_moment, UInt64 count) static inline Float64 apply(Float64 co_moment, Float64 left_m2, Float64 right_m2, UInt64 count)
{ {
if (count < 2) if (count < 2)
return 0.0; return 0.0;
@ -347,7 +376,7 @@ struct CovarPopImpl
{ {
static constexpr auto name = "covarPop"; static constexpr auto name = "covarPop";
static inline Float64 apply(Float64 co_moment, UInt64 count) static inline Float64 apply(Float64 co_moment, Float64 left_m2, Float64 right_m2, UInt64 count)
{ {
if (count < 2) if (count < 2)
return 0.0; return 0.0;
@ -356,6 +385,21 @@ struct CovarPopImpl
} }
}; };
/** Реализация функции corr.
*/
struct CorrImpl
{
static constexpr auto name = "corr";
static inline Float64 apply(Float64 co_moment, Float64 left_m2, Float64 right_m2, UInt64 count)
{
if (count < 2)
return 0.0;
else
return co_moment / sqrt(left_m2 * right_m2);
}
};
} }
template<typename T> template<typename T>
@ -376,4 +420,7 @@ using AggregateFunctionCovarSamp = AggregateFunctionCovariance<T, U, CovarSampIm
template<typename T, typename U> template<typename T, typename U>
using AggregateFunctionCovarPop = AggregateFunctionCovariance<T, U, CovarPopImpl>; using AggregateFunctionCovarPop = AggregateFunctionCovariance<T, U, CovarPopImpl>;
template<typename T, typename U>
using AggregateFunctionCorr = AggregateFunctionCovariance<T, U, CorrImpl, true>;
} }

View File

@ -617,6 +617,18 @@ AggregateFunctionPtr AggregateFunctionFactory::get(const String & name, const Da
return res; return res;
} }
else if (name == "corr")
{
if (argument_types.size() != 2)
throw Exception("Incorrect number of arguments for aggregate function " + name, ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
AggregateFunctionPtr res = createWithTwoNumericTypes<AggregateFunctionCorr>(*argument_types[0], *argument_types[1]);
if (!res)
throw Exception("Illegal types " + argument_types[0]->getName() + " and " + argument_types[1]->getName()
+ " of arguments for aggregate function " + name, ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
return res;
}
else if (recursion_level == 0 && name.size() > strlen("State") && !(strcmp(name.data() + name.size() - strlen("State"), "State"))) else if (recursion_level == 0 && name.size() > strlen("State") && !(strcmp(name.data() + name.size() - strlen("State"), "State")))
{ {
/// Для агрегатных функций вида aggState, где agg - имя другой агрегатной функции. /// Для агрегатных функций вида aggState, где agg - имя другой агрегатной функции.
@ -718,7 +730,8 @@ const AggregateFunctionFactory::FunctionNames & AggregateFunctionFactory::getFun
"stddevSamp", "stddevSamp",
"stddevPop", "stddevPop",
"covarSamp", "covarSamp",
"covarPop" "covarPop",
"corr"
}; };
return names; return names;