org.apache.drill.exec.physical.impl.common

Class HashTableTemplate

java.lang.Object

org.apache.drill.exec.physical.impl.common.HashTableTemplate

All Implemented Interfaces:

HashTable

public abstract class HashTableTemplate
extends Object
implements HashTable

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
class	HashTableTemplate.BatchHolder

Nested classes/interfaces inherited from interface org.apache.drill.exec.physical.impl.common.HashTable

HashTable.PutStatus

Field Summary

Fields

Modifier and Type	Field and Description
protected ClassGenerator<?>	cg
protected FragmentContext	context
static int	MAX_VARCHAR_SIZE

Fields inherited from interface org.apache.drill.exec.physical.impl.common.HashTable

BATCH_MASK, BATCH_SIZE, DEFAULT_LOAD_FACTOR, MAXIMUM_CAPACITY, TEMPLATE_DEFINITION

Constructor Summary

Constructors

Constructor and Description
HashTableTemplate()

Method Summary

Modifier and Type	Method and Description
void	clear() Frees all the direct memory consumed by the HashTable.
protected abstract void	doSetup(VectorContainer incomingBuild, RecordBatch incomingProbe)
void	enlargeEmptyHashTableIfNeeded(int newNum) Resize up the Hash Table if needed (to hold newNum entries)
long	getActualSize() The amount of direct memory consumed by the hash table.
int	getBuildHashCode(int incomingRowIdx) Return the Hash Value for the row in the Build incoming batch at index: (For Hash Aggregate there's no "Build" side -- only one batch - this one)
protected abstract int	getHashBuild(int incomingRowIdx, int seedValue)
protected abstract int	getHashProbe(int incomingRowIdx, int seedValue)
int	getProbeHashCode(int incomingRowIdx) Return the Hash Value for the row in the Probe incoming batch at index:
void	getStats(HashTableStats stats)
int	getTargetBatchRowCount()
protected HashTableTemplate.BatchHolder	injectMembers(HashTableTemplate.BatchHolder batchHolder)
boolean	isEmpty()
String	makeDebugString() Returns a message containing memory usage statistics.
protected HashTableTemplate.BatchHolder	newBatchHolder(int index, int newBatchHolderSize)
org.apache.commons.lang3.tuple.Pair<VectorContainer,Integer>	nextBatch()
int	numBuckets()
int	numResizing()
boolean	outputKeys(int batchIdx, VectorContainer outContainer, int numRecords) Retrieves the key columns and transfers them to the output container.
int	probeForKey(int incomingRowIdx, int hashCode) Return -1 if Probe-side key is not found in the (build-side) hash table.
HashTable.PutStatus	put(int incomingRowIdx, IndexPointer htIdxHolder, int hashCode, int targetBatchRowCount) put() uses the hash code (from gethashCode() above) to insert the key(s) from the incoming row into the hash table.
void	reset() Reinit the hash table to its original size, and clear up all its prior batch holder
void	setTargetBatchRowCount(int batchRowCount)
void	setup(HashTableConfig htConfig, BufferAllocator allocator, VectorContainer incomingBuild, RecordBatch incomingProbe, RecordBatch outgoing, VectorContainer htContainerOrig, FragmentContext context, ClassGenerator<?> cg) HashTable.setup(org.apache.drill.exec.physical.impl.common.HashTableConfig, org.apache.drill.exec.memory.BufferAllocator, org.apache.drill.exec.record.VectorContainer, org.apache.drill.exec.record.RecordBatch, org.apache.drill.exec.record.RecordBatch, org.apache.drill.exec.record.VectorContainer, org.apache.drill.exec.ops.FragmentContext, org.apache.drill.exec.expr.ClassGenerator<?>) must be called before anything can be done to the HashTable.
int	size()
void	updateBatches() Updates the incoming (build and probe side) value vectors references in the HashTableTemplate.BatchHolders.
void	updateIncoming(VectorContainer newIncoming, RecordBatch newIncomingProbe) Changes the incoming probe and build side batches, and then updates all the value vector references in the HashTableTemplate.BatchHolders.
void	updateInitialCapacity(int initialCapacity) Update the initial capacity for the hash table.

Methods inherited from class java.lang.Object

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

Field Detail

MAX_VARCHAR_SIZE

public static final int MAX_VARCHAR_SIZE

See Also:

Constant Field Values

context

protected FragmentContext context

cg

protected ClassGenerator<?> cg

Constructor Detail

HashTableTemplate

public HashTableTemplate()

Method Detail

setup

public void setup(HashTableConfig htConfig,
                  BufferAllocator allocator,
                  VectorContainer incomingBuild,
                  RecordBatch incomingProbe,
                  RecordBatch outgoing,
                  VectorContainer htContainerOrig,
                  FragmentContext context,
                  ClassGenerator<?> cg)

Description copied from interface: HashTable

HashTable.setup(org.apache.drill.exec.physical.impl.common.HashTableConfig, org.apache.drill.exec.memory.BufferAllocator, org.apache.drill.exec.record.VectorContainer, org.apache.drill.exec.record.RecordBatch, org.apache.drill.exec.record.RecordBatch, org.apache.drill.exec.record.VectorContainer, org.apache.drill.exec.ops.FragmentContext, org.apache.drill.exec.expr.ClassGenerator<?>) must be called before anything can be done to the HashTable.

Specified by:

setup in interface HashTable

updateInitialCapacity

public void updateInitialCapacity(int initialCapacity)

Description copied from interface: HashTable

Update the initial capacity for the hash table. This method will be removed after the key vectors are removed from the hash table. It is used to allocate HashTableTemplate.BatchHolders of appropriate size when the final size of the HashTable is known. Warning! Only call this method before you have inserted elements into the HashTable.

Specified by:

updateInitialCapacity in interface HashTable

Parameters:

initialCapacity - The new initial capacity to use.

updateBatches

public void updateBatches()
                   throws SchemaChangeException

Description copied from interface: HashTable

Updates the incoming (build and probe side) value vectors references in the HashTableTemplate.BatchHolders. This is useful on OK_NEW_SCHEMA (need to verify).

Specified by:

updateBatches in interface HashTable

Throws:

SchemaChangeException

numBuckets

public int numBuckets()

numResizing

public int numResizing()

size

public int size()

Specified by:

size in interface HashTable

getStats

public void getStats(HashTableStats stats)

Specified by:

getStats in interface HashTable

isEmpty

public boolean isEmpty()

Specified by:

isEmpty in interface HashTable

clear

public void clear()

Description copied from interface: HashTable

Frees all the direct memory consumed by the HashTable.

Specified by:

clear in interface HashTable

getBuildHashCode

public int getBuildHashCode(int incomingRowIdx)
                     throws SchemaChangeException

Return the Hash Value for the row in the Build incoming batch at index: (For Hash Aggregate there's no "Build" side -- only one batch - this one)

Specified by:

getBuildHashCode in interface HashTable

Parameters:

incomingRowIdx -

Returns:

Throws:

SchemaChangeException

getProbeHashCode

public int getProbeHashCode(int incomingRowIdx)
                     throws SchemaChangeException

Return the Hash Value for the row in the Probe incoming batch at index:

Specified by:

getProbeHashCode in interface HashTable

Parameters:

incomingRowIdx -

Returns:

Throws:

SchemaChangeException

put

public HashTable.PutStatus put(int incomingRowIdx,
                               IndexPointer htIdxHolder,
                               int hashCode,
                               int targetBatchRowCount)
                        throws SchemaChangeException,
                               RetryAfterSpillException

put() uses the hash code (from gethashCode() above) to insert the key(s) from the incoming row into the hash table. The code selects the bucket in the startIndices, then the keys are placed into the chained list - by storing the key values into a batch, and updating its "links" member. Last it modifies the index holder to the batch offset so that the caller can store the remaining parts of the row into a matching batch (outside the hash table). Returning

Specified by:

put in interface HashTable

Parameters:

incomingRowIdx - - position of the incoming row

htIdxHolder - - to return batch + batch-offset (for caller to manage a matching batch)

hashCode - - computed over the key(s) by calling getBuildHashCode()

Returns:

Status - the key(s) was ADDED or was already PRESENT

Throws:

SchemaChangeException

RetryAfterSpillException

probeForKey

public int probeForKey(int incomingRowIdx,
                       int hashCode)
                throws SchemaChangeException

Return -1 if Probe-side key is not found in the (build-side) hash table. Otherwise, return the global index of the key

Specified by:

probeForKey in interface HashTable

Parameters:

incomingRowIdx -

hashCode - - The hash code for the Probe-side key

Returns:

-1 if key is not found, else return the global index of the key

Throws:

SchemaChangeException

newBatchHolder

protected HashTableTemplate.BatchHolder newBatchHolder(int index,
                                                       int newBatchHolderSize)

injectMembers

protected HashTableTemplate.BatchHolder injectMembers(HashTableTemplate.BatchHolder batchHolder)

enlargeEmptyHashTableIfNeeded

public void enlargeEmptyHashTableIfNeeded(int newNum)

Resize up the Hash Table if needed (to hold newNum entries)

reset

public void reset()

Reinit the hash table to its original size, and clear up all its prior batch holder

Specified by:

reset in interface HashTable

updateIncoming

public void updateIncoming(VectorContainer newIncoming,
                           RecordBatch newIncomingProbe)

Description copied from interface: HashTable

Changes the incoming probe and build side batches, and then updates all the value vector references in the HashTableTemplate.BatchHolders.

Specified by:

updateIncoming in interface HashTable

Parameters:

newIncoming - The new build side batch.

newIncomingProbe - The new probe side batch.

outputKeys

public boolean outputKeys(int batchIdx,
                          VectorContainer outContainer,
                          int numRecords)

Description copied from interface: HashTable

Retrieves the key columns and transfers them to the output container. Note this operation removes the key columns from the HashTable.

Specified by:

outputKeys in interface HashTable

Parameters:

batchIdx - The index of a HashTableTemplate.BatchHolder in the HashTable.

outContainer - The destination container for the key columns.

numRecords - The number of key recorts to transfer.

Returns:

nextBatch

public org.apache.commons.lang3.tuple.Pair<VectorContainer,Integer> nextBatch()

Specified by:

nextBatch in interface HashTable

doSetup

protected abstract void doSetup(@Named(value="incomingBuild")
                                VectorContainer incomingBuild,
                                @Named(value="incomingProbe")
                                RecordBatch incomingProbe)
                         throws SchemaChangeException

Throws:

SchemaChangeException

getHashBuild

protected abstract int getHashBuild(@Named(value="incomingRowIdx")
                                    int incomingRowIdx,
                                    @Named(value="seedValue")
                                    int seedValue)
                             throws SchemaChangeException

Throws:

SchemaChangeException

getHashProbe

protected abstract int getHashProbe(@Named(value="incomingRowIdx")
                                    int incomingRowIdx,
                                    @Named(value="seedValue")
                                    int seedValue)
                             throws SchemaChangeException

Throws:

SchemaChangeException

getActualSize

public long getActualSize()

Description copied from interface: HashTable

The amount of direct memory consumed by the hash table.

Specified by:

getActualSize in interface HashTable

Returns:

makeDebugString

public String makeDebugString()

Description copied from interface: HashTable

Returns a message containing memory usage statistics. Intended to be used for printing debugging or error messages.

Specified by:

makeDebugString in interface HashTable

Returns:

A debug string.

setTargetBatchRowCount

public void setTargetBatchRowCount(int batchRowCount)

Specified by:

setTargetBatchRowCount in interface HashTable

getTargetBatchRowCount

public int getTargetBatchRowCount()

Specified by:

getTargetBatchRowCount in interface HashTable