com.bigdata.service

Class Split

java.lang.Object

com.bigdata.service.Split

public class Split
extends Object

Describes a "split" of keys for a batch operation. This is used in scale-out where the operation is parallelized across multiple index partitions. It is also used within a single index partition when an operation that has a lot of keys in its keys[] is parallelized over sub-key-ranges of that keys[].

Author:

Bryan Thompson

Field Summary

Fields

Modifier and Type	Field and Description
int	fromIndex Index of the first key in this split.
int	ntuples The #of keys in this split (toIndex - fromIndex).
IPartitionMetadata	pmd The index partition that spans the keys in this split.
int	toIndex Index of the first key NOT included in this split.

Constructor Summary

Constructors

Constructor and Description
Split(IPartitionMetadata pmd) Create a representation of a split point without specifying the from/to tuple index.
Split(IPartitionMetadata pmd, int fromIndex, int toIndex) Create a representation of a split point.

Method Summary

Methods

Modifier and Type	Method and Description
boolean	equals(Object o)
boolean	equals(Split o)
int	hashCode() Hash code is based on the IPartitionMetadata hash code if given (if is always present for scale-out client requests) and otherwise fromIndex (for example, when parallelizing operations within a single index per BLGZ-1537).
String	toString() Human friendly representation.

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Field Detail

pmd

public final IPartitionMetadata pmd

The index partition that spans the keys in this split.

fromIndex

public final int fromIndex

Index of the first key in this split.

toIndex

public final int toIndex

Index of the first key NOT included in this split.

ntuples

public final int ntuples

The #of keys in this split (toIndex - fromIndex).

Constructor Detail

Split

public Split(IPartitionMetadata pmd)

Create a representation of a split point without specifying the from/to tuple index.

Parameters:

pmd - The metadata for the index partition within which the keys in this split lie.

Split

public Split(IPartitionMetadata pmd,
     int fromIndex,
     int toIndex)

Create a representation of a split point.

Parameters:

pmd - The metadata for the index partition within which the keys in this split lie (optional and used only in scale-out).

fromIndex - The index of the first key that will enter that index partition (inclusive lower bound).

toIndex - The index of the first key that will NOT enter that index partition (exclusive upper bound).

Method Detail

hashCode

public int hashCode()

Hash code is based on the IPartitionMetadata hash code if given (if is always present for scale-out client requests) and otherwise fromIndex (for example, when parallelizing operations within a single index per BLGZ-1537).

Note: The historical hash code was just based on the IPartitionMetadata and would thrown out an NPE if there was no IPartitionMetadata. This was changed as part of BLZG-1537 to support cases where the IPartitionMetadata was not given (for index local parallelism).

Overrides:

hashCode in class Object

See Also:

(Schedule more IOs when loading data)

equals

public boolean equals(Object o)

Overrides:

equals in class Object

equals

public boolean equals(Split o)

toString

public String toString()

Human friendly representation.

Overrides:

toString in class Object