org.apache.drill.exec.record

Class RecordBatchSizer.ColumnSize

java.lang.Object

org.apache.drill.exec.record.RecordBatchSizer.ColumnSize

Enclosing class:

RecordBatchSizer

public class RecordBatchSizer.ColumnSize
extends Object

Column size information.

Field Summary

Fields

Modifier and Type	Field and Description
MaterializedField	metadata
String	prefix

Constructor Summary

Constructors

Constructor and Description
ColumnSize(ValueVector v, String prefix)

Method Summary

Modifier and Type	Method and Description
void	allocateVector(ValueVector vector, int recordCount)
void	buildVectorInitializer(VectorInitializer initializer) Add a single vector initializer to a collection for the entire batch.
int	getAllocSizePerEntry() This returns actual entry size if rowCount > 0 or allocation size otherwise.
float	getCardinality()
Map<String,RecordBatchSizer.ColumnSize>	getChildren()
int	getDataSizePerEntry() This is the average actual per entry data size in bytes.
int	getElementCount()
int	getNetSizePerEntry() This is the average per entry size of just pure data plus overhead of additional vectors we add on top like bits vector, offset vector etc.
int	getStdDataSizePerEntry() std pure data size per entry from Drill metadata, based on type.
int	getStdNetOrNetSizePerEntry() If there is an accurate std net size, that is returned.
int	getStdNetSizePerEntry() std net size per entry taking into account additional metadata vectors we add on top for variable length, cardinality etc.
int	getTotalDataSize() This is the total data size for the column, including children for map columns.
int	getTotalNetSize() This is the total net size for the column, including children for map columns.
int	getValueCount()
boolean	hasStdDataSize() Returns true if there is an accurate std size.
boolean	isComplex()
boolean	isRepeatedList()
boolean	isVariableWidth()
String	toString()

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Detail

prefix

public final String prefix

metadata

public final MaterializedField metadata

Constructor Detail

ColumnSize

public ColumnSize(ValueVector v,
                  String prefix)

Method Detail

hasStdDataSize

public boolean hasStdDataSize()

Returns true if there is an accurate std size. Otherwise it returns false.

Returns:

True if there is an accurate std size. Otherwise it returns false.

getStdDataSizePerEntry

public int getStdDataSizePerEntry()

std pure data size per entry from Drill metadata, based on type. Does not include metadata vector overhead we add for cardinality, variable length etc. For variable-width columns, we use 50 as std size for entry width. For repeated column, we assume repetition of 10.

getStdNetSizePerEntry

public int getStdNetSizePerEntry()

std net size per entry taking into account additional metadata vectors we add on top for variable length, cardinality etc. For variable-width columns, we use 50 as std data size for entry width. For repeated column, we assume repetition of 10.

getDataSizePerEntry

public int getDataSizePerEntry()

This is the average actual per entry data size in bytes. Does not include any overhead of metadata vectors. For repeated columns, it is average for the repeated array, not individual entry in the array.

getNetSizePerEntry

public int getNetSizePerEntry()

This is the average per entry size of just pure data plus overhead of additional vectors we add on top like bits vector, offset vector etc. This size is larger than the actual data size since this size includes per- column overhead for additional vectors we add for cardinality, variable length etc.

getAllocSizePerEntry

public int getAllocSizePerEntry()

This returns actual entry size if rowCount > 0 or allocation size otherwise. Use this for the cases when you might get empty batches with schema and you still need to do memory calculations based on just schema.

getStdNetOrNetSizePerEntry

public int getStdNetOrNetSizePerEntry()

If there is an accurate std net size, that is returned. Otherwise the net size is returned.

Returns:

If there is an accurate std net size, that is returned. Otherwise the net size is returned.

getTotalDataSize

public int getTotalDataSize()

This is the total data size for the column, including children for map columns. Does not include any overhead of metadata vectors.

getTotalNetSize

public int getTotalNetSize()

This is the total net size for the column, including children for map columns. Includes overhead of metadata vectors.

getValueCount

public int getValueCount()

getElementCount

public int getElementCount()

getCardinality

public float getCardinality()

isVariableWidth

public boolean isVariableWidth()

getChildren

public Map<String,RecordBatchSizer.ColumnSize> getChildren()

isComplex

public boolean isComplex()

isRepeatedList

public boolean isRepeatedList()

allocateVector

public void allocateVector(ValueVector vector,
                           int recordCount)

toString

public String toString()

Overrides:

toString in class Object

buildVectorInitializer

public void buildVectorInitializer(VectorInitializer initializer)

Add a single vector initializer to a collection for the entire batch. Uses the observed column size information to predict the size needed when allocating a new vector for the same data. Adds a hint only for variable-width or repeated types; no extra information is needed for fixed width, non-repeated columns.

Parameters:

initializer - the vector initializer to hold the hints for this column