Package org.apache.drill.exec.work.filter

Class BloomFilter

java.lang.Object

org.apache.drill.exec.work.filter.BloomFilter

public class BloomFilter
extends Object

According to Putze et al.'s "Cache-, Hash- and Space-Efficient BloomFilter Filters", see this paper for details, the main theory is to construct tiny bucket bloom filters which benefit to the cpu cache and SIMD opcode.

Constructor Summary

Constructors

Constructor	Description
BloomFilter(int ndv, double fpp, BufferAllocator bufferAllocator)
BloomFilter(int numBytes, BufferAllocator bufferAllocator)
BloomFilter(DrillBuf byteBuf)

Method Summary

Modifier and Type	Method	Description
static int	adjustByteSize(int numBytes)
boolean	find(long hash)	Determine whether an element is set or not.
DrillBuf	getContent()
void	insert(long hash)	Add an element's hash value to this bloom filter.
static int	optimalNumOfBytes(long ndv, double fpp)	Calculate optimal size according to the number of distinct values and false positive probability.
void	or(BloomFilter other)	Merge this bloom filter with other one

Methods inherited from class java.lang.Object

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

Constructor Details

BloomFilter

public BloomFilter(int numBytes,
 BufferAllocator bufferAllocator)

BloomFilter

public BloomFilter(int ndv,
 double fpp,
 BufferAllocator bufferAllocator)

BloomFilter

public BloomFilter(DrillBuf byteBuf)

Method Details

adjustByteSize

public static int adjustByteSize(int numBytes)

insert

public void insert(long hash)

Add an element's hash value to this bloom filter.

Parameters:

hash - hash result of element.

find

public boolean find(long hash)

Determine whether an element is set or not.

Parameters:

hash - the hash value of element.

Returns:

false if the element is not set, true if the element is probably set.

or

public void or(BloomFilter other)

Merge this bloom filter with other one

Parameters:

other - other bloom filter

optimalNumOfBytes

public static int optimalNumOfBytes(long ndv,
 double fpp)

Calculate optimal size according to the number of distinct values and false positive probability. See http://en.wikipedia.org/wiki/Bloom_filter#Probability_of_false_positives for the formula.

Parameters:

ndv - The number of distinct values.

fpp - The false positive probability.

Returns:

optimal number of bytes of given ndv and fpp.

getContent

public DrillBuf getContent()