com.bigdata.btree.proc

Class AbstractKeyArrayIndexProcedure<T>

java.lang.Object

com.bigdata.btree.proc.AbstractIndexProcedure<T>

com.bigdata.btree.proc.AbstractKeyArrayIndexProcedure<T>

All Implemented Interfaces:

IIndexProcedure<T>, IKeyArrayIndexProcedure<T>, IReadOnly, Externalizable, Serializable

Direct Known Subclasses:

BatchContains, BatchInsert, BatchLookup, BatchPutIfAbsent, BatchRemove, BlobsWriteProc, Id2TermWriteProc, JustIndexWriteProc, SPOIndexWriteProc, Term2IdWriteProc, TextIndexWriteProc

public abstract class AbstractKeyArrayIndexProcedure<T>
extends AbstractIndexProcedure<T>
implements IKeyArrayIndexProcedure<T>, Externalizable

Abstract base class supports compact serialization and compression for remote IKeyArrayIndexProcedure execution (procedures may be executed on a local index, but they are only (de-)serialized when executed on a remote index).

Author:

Bryan Thompson

See Also:

(Schedule more IOs when loading data), Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	AbstractKeyArrayIndexProcedure.ResultBitBuffer A class useful for sending a logical boolean[] back from a remote procedure call.
static class	AbstractKeyArrayIndexProcedure.ResultBitBufferCounter Counts the #of true bits in the AbstractKeyArrayIndexProcedure.ResultBitBuffer(s).
static class	AbstractKeyArrayIndexProcedure.ResultBitBufferHandler Knows how to aggregate AbstractKeyArrayIndexProcedure.ResultBitBuffer objects.
static class	AbstractKeyArrayIndexProcedure.ResultBuffer A class useful for sending some kinds of data back from a remote procedure call (those readily expressed as a byte[][]).
static class	AbstractKeyArrayIndexProcedure.ResultBufferHandler Knows how to aggregate AbstractKeyArrayIndexProcedure.ResultBuffer objects.

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	AbstractKeyArrayIndexProcedure() De-serialization constructor.
protected	AbstractKeyArrayIndexProcedure(IRabaCoder keysCoder, IRabaCoder valsCoder, int fromIndex, int toIndex, byte[][] keys, byte[][] vals)

Method Summary

Methods

Modifier and Type	Method and Description
T	apply(IIndex ndx) Applies the logic of the procedure.
protected abstract T	applyOnce(IIndex ndx, IRaba keys, IRaba vals) Apply the procedure to the specified key range of the index.
IRaba	getKeys() The keys.
protected IRabaCoder	getKeysCoder() The object used to (de-)code the keys when they are sent to the remote service.
IRaba	getValues() The values.
protected IRabaCoder	getValuesCoder() The object used to (de-)code the values when they are sent to the remote service.
protected abstract IResultHandler<T,T>	newAggregator() Return an IResultHandler that will be used to combine the results if the index procedure is parallelized against a local index (including a scale-out shard).
void	readExternal(ObjectInput in)
protected void	readMetadata(ObjectInput in) Reads metadata written by writeMetadata(ObjectOutput).
void	writeExternal(ObjectOutput out)
protected void	writeMetadata(ObjectOutput out) Writes metadata (not the keys or values, but just other metadata used by the procedure).

Methods inherited from class java.lang.Object

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

Methods inherited from interface com.bigdata.journal.IReadOnly

isReadOnly

Constructor Detail

AbstractKeyArrayIndexProcedure

protected AbstractKeyArrayIndexProcedure()

De-serialization constructor.

AbstractKeyArrayIndexProcedure

protected AbstractKeyArrayIndexProcedure(IRabaCoder keysCoder,
                              IRabaCoder valsCoder,
                              int fromIndex,
                              int toIndex,
                              byte[][] keys,
                              byte[][] vals)

Parameters:

keysCoder - The object used to serialize the keys.

valsCoder - The object used to serialize the vals (optional IFF vals is null).

fromIndex - The index of the first key in keys to be processed (inclusive).

toIndex - The index of the last key in keys to be processed.

keys - The keys (unsigned variable length byte[]s) MUST be in sorted order (the logic to split procedures across partitioned indices depends on this, plus ordered reads and writes on indices are MUCH more efficient).

vals - The values (optional, must be co-indexed with keys when non-null).

Method Detail

getKeysCoder

protected IRabaCoder getKeysCoder()

The object used to (de-)code the keys when they are sent to the remote service.

getValuesCoder

protected IRabaCoder getValuesCoder()

The object used to (de-)code the values when they are sent to the remote service.

getKeys

public final IRaba getKeys()

Description copied from interface: IKeyArrayIndexProcedure

The keys.

Specified by:

getKeys in interface IKeyArrayIndexProcedure<T>

Returns:

The keys and never null.

getValues

public final IRaba getValues()

Description copied from interface: IKeyArrayIndexProcedure

The values.

Specified by:

getValues in interface IKeyArrayIndexProcedure<T>

Returns:

The values -or- null if no values were associated with the IIndexProcedure.

newAggregator

protected abstract IResultHandler<T,T> newAggregator()

Return an IResultHandler that will be used to combine the results if the index procedure is parallelized against a local index (including a scale-out shard). If a null is returned, then the index procedure WILL NOT be parallelized against the local index. To parallelize index procedures that do not return anything against a local index, just use NopAggregator. A non-null value will permit both index local parallelization of the index procedure and (in scale-out) parallelization of the index procedure across the shards as well. In order to be parallelized, the index procedure must also be marked as IParallelizableIndexProcedure.

Returns:

The IResultHandler -or- null

See Also:

NopAggregator, IParallelizableIndexProcedure, (Schedule more IOs when loading data)

apply

public final T apply(IIndex ndx)

Applies the logic of the procedure.

Note: For invocations where the IRaba.size() of the keys is large, this class breaks down the keys into a multiple key ranges to parallelize the work. If the procedure is read-only, then we can trivially parallelize the operation. When the procedure is read-write, a prefetch pattern is used to ensure that the index pages are in cache and then work is handed off to a single thread that does the actual work while obeying the single-threaded for writer constraint on the index.

Specified by:

apply in interface IIndexProcedure<T>

Parameters:

ndx - The index.

Returns:

The result, which is entirely defined by the procedure implementation and which MAY be null. In general, this MUST be Serializable since it may have to pass across a network interface.

See Also:

(Schedule more IOs when loading data)

applyOnce

protected abstract T applyOnce(IIndex ndx,
          IRaba keys,
          IRaba vals)

Apply the procedure to the specified key range of the index.

Parameters:

ndx - The index.

Returns:

readExternal

public final void readExternal(ObjectInput in)
                        throws IOException,
                               ClassNotFoundException

Specified by:

readExternal in interface Externalizable

Throws:

IOException

ClassNotFoundException

writeExternal

public final void writeExternal(ObjectOutput out)
                         throws IOException

Specified by:

writeExternal in interface Externalizable

Throws:

IOException

readMetadata

protected void readMetadata(ObjectInput in)
                     throws IOException,
                            ClassNotFoundException

Reads metadata written by writeMetadata(ObjectOutput).

Parameters:

in -

Throws:

IOException

ClassNotFoundException

writeMetadata

protected void writeMetadata(ObjectOutput out)
                      throws IOException

Writes metadata (not the keys or values, but just other metadata used by the procedure).

The default implementation writes out the getKeysCoder() and the getValuesCoder().

Parameters:

out -

Throws:

IOException