org.apache.datasketches

Class HashOperations

java.lang.Object

org.apache.datasketches.HashOperations

public final class HashOperations
extends Object

Helper class for the common hash table methods.

Author:

Lee Rhodes, Kevin Lang

Field Summary

Fields

Modifier and Type	Field and Description
static int	STRIDE_MASK The stride mask for the Open Address, Double Hashing (OADH) hash table algorithm.

Method Summary

Modifier and Type	Method and Description
static void	checkHashCorruption(long hash)
static void	checkThetaCorruption(long thetaLong)
static boolean	continueCondition(long thetaLong, long hash) Return true (continue) if hash is greater than or equal to thetaLong, or if hash == 0, or if hash == Long.MAX_VALUE.
static int	count(long[] srcArr, long thetaLong) Counts the cardinality of the given source array.
static int	countPart(long[] srcArr, int lgArrLongs, long thetaLong) Counts the cardinality of the first Log2 values of the given source array.
static int	fastHashInsertOnly(org.apache.datasketches.memory.WritableMemory wmem, int lgArrLongs, long hash, int memOffsetBytes) This is a classical Knuth-style Open Addressing, Double Hash insert scheme, but inserts values directly into a Memory.
static int	fastHashSearchOrInsert(org.apache.datasketches.memory.WritableMemory wmem, int lgArrLongs, long hash, int memOffsetBytes) This is a classical Knuth-style Open Addressing, Double Hash insert scheme, but inserts values directly into a Memory.
static int	hashArrayInsert(long[] srcArr, long[] hashTable, int lgArrLongs, long thetaLong) Inserts the given long array into the given hash table array of the target size, removes any negative input values, ignores duplicates and counts the values inserted.
static int	hashInsertOnly(long[] hashTable, int lgArrLongs, long hash) This is a classical Knuth-style Open Addressing, Double Hash insert scheme for on-heap.
static int	hashSearch(long[] hashTable, int lgArrLongs, long hash) This is a classical Knuth-style Open Addressing, Double Hash search scheme for on-heap.
static int	hashSearch(org.apache.datasketches.memory.Memory mem, int lgArrLongs, long hash, int memOffsetBytes) This is a classical Knuth-style Open Addressing, Double Hash search scheme for off-heap.
static int	hashSearchOrInsert(long[] hashTable, int lgArrLongs, long hash) This is a classical Knuth-style Open Addressing, Double Hash insert scheme for on-heap.

Methods inherited from class java.lang.Object

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

Field Detail

STRIDE_MASK

public static final int STRIDE_MASK

The stride mask for the Open Address, Double Hashing (OADH) hash table algorithm.

See Also:

Constant Field Values

Method Detail

countPart

public static int countPart(long[] srcArr,
                            int lgArrLongs,
                            long thetaLong)

Counts the cardinality of the first Log2 values of the given source array.

Parameters:

srcArr - the given source array

lgArrLongs - See lgArrLongs

thetaLong - See Theta Long

Returns:

the cardinality

count

public static int count(long[] srcArr,
                        long thetaLong)

Counts the cardinality of the given source array.

Parameters:

srcArr - the given source array

thetaLong - See Theta Long

Returns:

the cardinality

hashSearch

public static int hashSearch(long[] hashTable,
                             int lgArrLongs,
                             long hash)

This is a classical Knuth-style Open Addressing, Double Hash search scheme for on-heap. Returns the index if found, -1 if not found.

Parameters:

hashTable - The hash table to search. Must be a power of 2 in size.

lgArrLongs - See lgArrLongs. lgArrLongs ≤ log2(hashTable.length).

hash - A hash value to search for. Must not be zero.

Returns:

Current probe index if found, -1 if not found.

hashInsertOnly

public static int hashInsertOnly(long[] hashTable,
                                 int lgArrLongs,
                                 long hash)

This is a classical Knuth-style Open Addressing, Double Hash insert scheme for on-heap. This method assumes that the input hash is not a duplicate. Useful for rebuilding tables to avoid unnecessary comparisons. Returns the index of insertion, which is always positive or zero. Throws an exception if the table is full with no empty slot. Throws an exception if table has no empty slot.

Parameters:

hashTable - the hash table to insert into.

lgArrLongs - See lgArrLongs. lgArrLongs ≤ log2(hashTable.length).

hash - value that must not be zero and will be inserted into the array into an empty slot.

Returns:

index of insertion. Always positive or zero.

hashSearchOrInsert

public static int hashSearchOrInsert(long[] hashTable,
                                     int lgArrLongs,
                                     long hash)

This is a classical Knuth-style Open Addressing, Double Hash insert scheme for on-heap. Returns index ≥ 0 if found (duplicate); < 0 if inserted, inserted at -(index + 1). Throws an exception if the value is not found and table has no empty slot.

Parameters:

hashTable - the hash table to insert into.

lgArrLongs - See lgArrLongs. lgArrLongs ≤ log2(hashTable.length).

hash - hash value that must not be zero and if not a duplicate will be inserted into the array into an empty slot

Returns:

index ≥ 0 if found (duplicate); < 0 if inserted, inserted at -(index + 1).

hashArrayInsert

public static int hashArrayInsert(long[] srcArr,
                                  long[] hashTable,
                                  int lgArrLongs,
                                  long thetaLong)

Inserts the given long array into the given hash table array of the target size, removes any negative input values, ignores duplicates and counts the values inserted. The given hash table may have values, but they must have been inserted by this method or one of the other OADH insert methods in this class and they may not be dirty. This method performs additional checks against potentially invalid hash values or theta values. Returns the count of values actually inserted.

Parameters:

srcArr - the source hash array to be potentially inserted

hashTable - The correctly sized target hash table that must be a power of two.

lgArrLongs - See lgArrLongs. lgArrLongs ≤ log2(hashTable.length).

thetaLong - must greater than zero See Theta Long

Returns:

the count of values actually inserted

hashSearch

public static int hashSearch(org.apache.datasketches.memory.Memory mem,
                             int lgArrLongs,
                             long hash,
                             int memOffsetBytes)

This is a classical Knuth-style Open Addressing, Double Hash search scheme for off-heap. Returns the index if found, -1 if not found.

Parameters:

mem - The Memory hash table to search.

lgArrLongs - See lgArrLongs. lgArrLongs ≤ log2(hashTable.length).

hash - A hash value to search for. Must not be zero.

memOffsetBytes - offset in the memory where the hash array starts

Returns:

index if found, -1 if not found.

fastHashInsertOnly

public static int fastHashInsertOnly(org.apache.datasketches.memory.WritableMemory wmem,
                                     int lgArrLongs,
                                     long hash,
                                     int memOffsetBytes)

This is a classical Knuth-style Open Addressing, Double Hash insert scheme, but inserts values directly into a Memory. This method assumes that the input hash is not a duplicate. Useful for rebuilding tables to avoid unnecessary comparisons. Returns the index of insertion, which is always positive or zero. Throws an exception if table has no empty slot.

Parameters:

wmem - The writable memory

lgArrLongs - See lgArrLongs. lgArrLongs ≤ log2(hashTable.length).

hash - value that must not be zero and will be inserted into the array into an empty slot.

memOffsetBytes - offset in the memory where the hash array starts

Returns:

index of insertion. Always positive or zero.

fastHashSearchOrInsert

public static int fastHashSearchOrInsert(org.apache.datasketches.memory.WritableMemory wmem,
                                         int lgArrLongs,
                                         long hash,
                                         int memOffsetBytes)

This is a classical Knuth-style Open Addressing, Double Hash insert scheme, but inserts values directly into a Memory. Returns index ≥ 0 if found (duplicate); < 0 if inserted, inserted at -(index + 1). Throws an exception if the value is not found and table has no empty slot.

Parameters:

wmem - the WritableMemory

lgArrLongs - See lgArrLongs. lgArrLongs ≤ log2(hashTable.length).

hash - A hash value that must not be zero and if not a duplicate will be inserted into the array into an empty slot.

memOffsetBytes - offset in the memory where the hash array starts

Returns:

index ≥ 0 if found (duplicate); < 0 if inserted, inserted at -(index + 1).

checkThetaCorruption

public static void checkThetaCorruption(long thetaLong)

Parameters:

thetaLong - must be greater than zero otherwise throws an exception. See Theta Long

checkHashCorruption

public static void checkHashCorruption(long hash)

Parameters:

hash - must be greater than -1 otherwise throws an exception. Note a hash of zero is normally ignored, but a negative hash is never allowed.

continueCondition

public static boolean continueCondition(long thetaLong,
                                        long hash)

Return true (continue) if hash is greater than or equal to thetaLong, or if hash == 0, or if hash == Long.MAX_VALUE.

Parameters:

thetaLong - must be greater than the hash value See Theta Long

hash - must be less than thetaLong and not less than or equal to zero.

Returns:

true (continue) if hash is greater than or equal to thetaLong, or if hash == 0, or if hash == Long.MAX_VALUE.