anonymous_namespace{cost_model.cc} Namespace Reference

Classes
class	AggregateRowEstimator
	This class finds disjoint sets of aggregation terms that form prefixes of some non-hash index, and makes row estimates for those sets based on index metadata. More...

Typedefs
using	TermArray = Bounds_checked_array< const Item *const >
	Array of aggregation terms. More...

Functions
double	EstimateAggregateRows (THD *thd, const AccessPath *child, const Query_block *query_block, bool rollup)
	Estimate the row count for an aggregate operation (including ROLLUP rows for GROUP BY ... WITH ROLLUP). More...
TermArray	GetAggregationTerms (const JOIN &join)
double	EstimateDistinctRowsFromStatistics (THD *thd, TermArray terms, double child_rows)
	Estimate the number of distinct tuples in the projection defined by 'terms'. More...
template<typename FunctionLow , typename FunctionHigh >
double	SmoothTransition (FunctionLow function_low, FunctionHigh function_high, double lower_limit, double upper_limit, double argument)
	For a function f(x) such that: f(x) = g(x) for x<=l f(x) = h(x) for x>l. More...
double	EstimateRollupRowsPrimitively (double aggregate_rows, size_t grouping_expressions)
	Do a cheap rollup row estimate for small result sets. More...
double	EstimateRollupRowsAdvanced (THD *thd, double aggregate_rows, TermArray terms)
	Do more precise rollup row estimate for larger result sets. More...

Typedef Documentation

TermArray

using anonymous_namespace{cost_model.cc}::TermArray = typedef Bounds_checked_array<const Item *const>

Array of aggregation terms.

Function Documentation

EstimateAggregateRows()

double anonymous_namespace cost_model anonymous_namespace{cost_model.cc}::cc::EstimateAggregateRows	(	THD *	thd,
		const AccessPath *	child,
		const Query_block *	query_block,
		bool	rollup
	)

Estimate the row count for an aggregate operation (including ROLLUP rows for GROUP BY ... WITH ROLLUP).

Parameters

thd	Current thread.
child	The input to the aggregate path.
query_block	The query block to which the aggregation belongs.
rollup	True if we should add rollup rows to the estimate.

Returns

The row estimate.

EstimateDistinctRowsFromStatistics()

double anonymous_namespace{cost_model.cc}::EstimateDistinctRowsFromStatistics	(	THD *	thd,
		TermArray	terms,
		double	child_rows
	)

Estimate the number of distinct tuples in the projection defined by 'terms'.

We use the following data to make a row estimate, in that priority:

1. (Non-hash) indexes where the terms form some prefix of the index key. The handler can give good estimates for these.
2. Statistics from secondary engine or histograms for terms that are fields. Both can give an estimate of the number of unique values. (Statistics from secondary engine is preferred if available.)
3. The table size (in rows) for terms that are fields without histograms. (If we have "SELECT ... FROM t1 JOIN t2 GROUP BY t2.f1", there cannot be more results rows than there are rows in t2.) We also make the pragmatic assumption that that field values are not unique, and therefore make a row estimate somewhat lower than the table row count.
4. In the remaining cases we make an estimate based on the input row estimate. This is based on two assumptions: a) There will be fewer output rows than input rows, as one rarely aggregates on a set of terms that are unique for each row, b) The more terms there are, the more output rows one can expect.

We may need to combine multiple estimates into one. As an example, assume that we aggregate on three fields: f1, f2 and f3. There is and index where f1, f2 are a key prefix, and we have a histogram on f3. Then we could make good estimates for "GROUP BY f1,f2" or "GROUP BY f3". But how do we combine these into an estimate for "GROUP BY f1,f2,f3"? If f3 and f1,f2 are uncorrelated, then we should multiply the individual estimates. But if f3 is functionally dependent on f1,f2 (or vice versa), we should pick the larger of the two estimates.

Since we do not know if these fields are correlated or not, we multiply the individual estimates and then multiply with a damping factor. The damping factor is a function of the number of estimates (two in the example above). That way, we get a combined estimate that falls between the two extremes of functional dependence and no correlation.

Parameters

thd	Current thread.
terms	The terms for which we estimate the number of distinct combinations.
child_rows	The row estimate for the input path.

Returns

The row estimate for the aggregate operation.

EstimateRollupRowsAdvanced()

double anonymous_namespace{cost_model.cc}::EstimateRollupRowsAdvanced	(	THD *	thd,
		double	aggregate_rows,
		TermArray	terms
	)

Do more precise rollup row estimate for larger result sets.

If we have ROLLUP, there will be additional rollup rows. If we group on N terms T1..TN, we assume that the number of rollup rows will be:

1 + CARD(T1) + CARD(T1,T2) +...CARD(T1...T(N-1))

were CARD(T1...TX) is a row estimate for aggregating on T1..TX.

Parameters

thd	Current thread.
aggregate_rows	Number of rows after aggregation.
terms	The group-by terms.

Returns

Estimated number of rollup rows.

EstimateRollupRowsPrimitively()

double anonymous_namespace{cost_model.cc}::EstimateRollupRowsPrimitively	(	double	aggregate_rows,
		size_t	grouping_expressions
	)

Do a cheap rollup row estimate for small result sets.

If we group on n terms and expect k rows in total (before rollup), we make the simplifying assumption that each term has k^(1/n) distinct values, and that all terms are uncorrelated from each other. Then the number of rollup rows can be expressed as the sum of a finite geometric series:

1 + m+ m^2+m^3...m^(n-1)

where m = k^(1/n).

Parameters

aggregate_rows	Number of rows after aggregation.
grouping_expressions	Number of terms that we aggregated on.

Returns

Estimated number of rollup rows.

GetAggregationTerms()

TermArray anonymous_namespace{cost_model.cc}::GetAggregationTerms

(

const JOIN &

join

)

SmoothTransition()

template<typename FunctionLow , typename FunctionHigh >

double anonymous_namespace{cost_model.cc}::SmoothTransition	(	FunctionLow	function_low,
		FunctionHigh	function_high,
		double	lower_limit,
		double	upper_limit,
		double	argument
	)

For a function f(x) such that: f(x) = g(x) for x<=l f(x) = h(x) for x>l.

tweak f(x) so that it is continuous at l even if g(l) != h(l). We obtain this by doing a gradual transition between g(x) and h(x) in an interval [l, l+k] for some constant k.

Parameters

function_low	g(x)
function_high	h(x)
lower_limit	l
upper_limit	l+k
argument	x (for f(x))

Returns

Tweaked f(x).