Inheritance diagram for duckdb::ReservoirSample:

Collaboration diagram for duckdb::ReservoirSample:

[legend]

Public Member Functions
	ReservoirSample (Allocator &allocator, idx_t sample_count, int64_t seed=1)

	ReservoirSample (idx_t sample_count, unique_ptr< ReservoirChunk >=nullptr)

void	EvictOverBudgetSamples ()
	methods used to help with serializing and deserializing

void	ExpandSerializedSample ()

SamplingState	GetSamplingState () const

void	Vacuum ()

void	ConvertToReservoirSample ()
	Transform To sample based on reservoir sampling paper.

template<typename T >
T	GetReservoirChunkCapacity () const
	Get the capactiy of the data chunk reserved for storing samples.

unique_ptr< BlockingSample >	Copy () const override

idx_t	FillReservoir (DataChunk &chunk)
	create the first chunk called by AddToReservoir()

void	AddToReservoir (DataChunk &input) override
	Add a chunk of data to the sample.

void	Merge (unique_ptr< BlockingSample > other)
	Merge two Reservoir Samples. Other must be a reservoir sample.

void	ShuffleSel (SelectionVector &sel, idx_t range, idx_t size) const

void	UpdateSampleAppend (DataChunk &this_, DataChunk &other, SelectionVector &other_sel, idx_t append_count) const

idx_t	GetTuplesSeen () const

idx_t	NumSamplesCollected () const

idx_t	GetActiveSampleCount () const

DataChunk &	Chunk ()

unique_ptr< DataChunk >	GetChunk () override

void	Destroy () override

void	Finalize () override

void	Verify ()

idx_t	GetSampleCount ()

SelectionVectorHelper	GetReplacementIndexes (idx_t sample_chunk_offset, idx_t theoretical_chunk_length)
	Get indexes from current sample that can be replaced.

void	Serialize (Serializer &serializer) const override

Public Member Functions inherited from duckdb::BlockingSample
	BlockingSample (int64_t seed=-1)

std::pair< double, idx_t >	PopFromWeightQueue ()
	Helper functions needed to merge two reservoirs while respecting weights of sampled rows.

double	GetMinWeightThreshold ()

idx_t	GetPriorityQueueSize ()

template<class TARGET >
TARGET &	Cast ()

template<class TARGET >
const TARGET &	Cast () const

Static Public Member Functions
static bool	ValidSampleType (const LogicalType &type)

static unique_ptr< BlockingSample >	Deserialize (Deserializer &deserializer)

Static Public Member Functions inherited from duckdb::BlockingSample
static unique_ptr< BlockingSample >	Deserialize (Deserializer &deserializer)

Static Public Attributes
static constexpr const SampleType	TYPE = SampleType::RESERVOIR_SAMPLE

static constexpr idx_t	FIXED_SAMPLE_SIZE_MULTIPLIER = 10

static constexpr idx_t	FAST_TO_SLOW_THRESHOLD = MinValue<idx_t>(STANDARD_VECTOR_SIZE, 60)

static constexpr double	SAVE_PERCENTAGE = 0.01

Static Public Attributes inherited from duckdb::BlockingSample
static constexpr const SampleType	TYPE = SampleType::BLOCKING_SAMPLE

Private Member Functions
void	NormalizeWeights ()

SelectionVectorHelper	GetReplacementIndexesSlow (const idx_t sample_chunk_offset, const idx_t chunk_length)

SelectionVectorHelper	GetReplacementIndexesFast (const idx_t sample_chunk_offset, const idx_t chunk_length)

void	SimpleMerge (ReservoirSample &other)

void	WeightedMerge (ReservoirSample &other_sample)

unique_ptr< ReservoirChunk >	CreateNewSampleChunk (vector< LogicalType > &types, idx_t size) const

vector< uint32_t >	GetRandomizedVector (uint32_t range, uint32_t size) const

Private Attributes
idx_t	sample_count

Allocator &	allocator

unique_ptr< ReservoirChunk >	reservoir_chunk

bool	stats_sample

SelectionVector	sel

idx_t	sel_size

Additional Inherited Members
Public Attributes inherited from duckdb::BlockingSample
unique_ptr< BaseReservoirSampling >	base_reservoir_sample

SampleType	type
	The sample type.

bool	destroyed
	has the sample been destroyed due to updates to the referenced table

Member Function Documentation

◆ Vacuum()

void duckdb::ReservoirSample::Vacuum ( )

Vacuum the Reservoir Sample so it throws away tuples that are not in the reservoir weights or in the selection vector

◆ Copy()

unique_ptr< BlockingSample > duckdb::ReservoirSample::Copy ( ) const

overridevirtual

If for_serialization=true then the sample_chunk is not padded with extra spaces for future sampling values

Implements duckdb::BlockingSample.

◆ AddToReservoir()

void duckdb::ReservoirSample::AddToReservoir ( DataChunk & input )

overridevirtual

Add a chunk of data to the sample.

Implements duckdb::BlockingSample.

◆ UpdateSampleAppend()

void duckdb::ReservoirSample::UpdateSampleAppend	(	DataChunk &	this_,
		DataChunk &	other,
		SelectionVector &	other_sel,
		idx_t	append_count
	)		const

Update the sample by pushing new sample rows to the end of the sample_chunk. The new sample rows are the tuples rows resulting from applying sel to other

◆ GetChunk()

unique_ptr< DataChunk > duckdb::ReservoirSample::GetChunk ( )

overridevirtual

Fetches a chunk from the sample. Note that this method is destructive and should only be used after the sample is completely built.

Implements duckdb::BlockingSample.

◆ Destroy()

void duckdb::ReservoirSample::Destroy ( )

overridevirtual

Reimplemented from duckdb::BlockingSample.

◆ Finalize()

void duckdb::ReservoirSample::Finalize ( )

overridevirtual

Implements duckdb::BlockingSample.

◆ Serialize()

void duckdb::ReservoirSample::Serialize ( Serializer & serializer ) const

overridevirtual

Reimplemented from duckdb::BlockingSample.

The documentation for this class was generated from the following file:

external/duckdb/duckdb.hpp

Public Member Functions

Static Public Member Functions

Static Public Attributes

Private Member Functions

Private Attributes

Additional Inherited Members

Member Function Documentation

◆ Vacuum()

◆ Copy()

◆ AddToReservoir()

◆ UpdateSampleAppend()

◆ GetChunk()

◆ Destroy()

◆ Finalize()

◆ Serialize()