org.apache.hadoop.hbase.util

Class RegionSplitCalculator<R extends KeyRange>

java.lang.Object

org.apache.hadoop.hbase.util.RegionSplitCalculator<R>

Type Parameters:

R -

@InterfaceAudience.Private
public class RegionSplitCalculator<R extends KeyRange>
extends Object

This is a generic region split calculator. It requires Ranges that provide start, end, and a comparator. It works in two phases -- the first adds ranges and rejects backwards ranges. Then one calls calcRegions to generate the multimap that has a start split key as a key and possibly multiple Ranges as members. To traverse, one normally would get the split set, and iterate through the calcRegions. Normal regions would have only one entry, holes would have zero, and any overlaps would have multiple entries. The interface is a bit cumbersome currently but is exposed this way so that clients can choose how to iterate through the region splits.

Field Summary

Fields

Modifier and Type	Field and Description
static Comparator<byte[]>	BYTES_COMPARATOR
private static byte[]	ENDKEY SPECIAL CASE
private static org.slf4j.Logger	LOG
private Comparator<R>	rangeCmp
private TreeSet<byte[]>	splits This contains a sorted set of all the possible split points Invariant: once populated this has 0 entries if empty or at most n+1 values where n == number of added ranges.
private org.apache.hbase.thirdparty.com.google.common.collect.Multimap<byte[],R>	starts This is a map from start key to regions with the same start key.

Constructor Summary

Constructors

Constructor and Description
RegionSplitCalculator(Comparator<R> cmp)

Method Summary

Modifier and Type	Method and Description
boolean	add(R range) Adds an edge to the split calculator
org.apache.hbase.thirdparty.com.google.common.collect.Multimap<byte[],R>	calcCoverage() Generates a coverage multimap from split key to Regions that start with the split key.
static <R extends KeyRange> List<R>	findBigRanges(Collection<R> bigOverlap, int count) Find specified number of top ranges in a big overlap group.
TreeSet<byte[]>	getSplits()
org.apache.hbase.thirdparty.com.google.common.collect.Multimap<byte[],R>	getStarts()
private static <R extends KeyRange> byte[]	specialEndKey(R range) SPECIAL CASE wrapper for empty end key

Methods inherited from class java.lang.Object

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

Field Detail

LOG

private static final org.slf4j.Logger LOG

rangeCmp

private final Comparator<R extends KeyRange> rangeCmp

splits

private final TreeSet<byte[]> splits

This contains a sorted set of all the possible split points Invariant: once populated this has 0 entries if empty or at most n+1 values where n == number of added ranges.

starts

private final org.apache.hbase.thirdparty.com.google.common.collect.Multimap<byte[],R extends KeyRange> starts

This is a map from start key to regions with the same start key. Invariant: This always have n values in total

ENDKEY

private static final byte[] ENDKEY

SPECIAL CASE

BYTES_COMPARATOR

public static final Comparator<byte[]> BYTES_COMPARATOR

Constructor Detail

RegionSplitCalculator

public RegionSplitCalculator(Comparator<R> cmp)

Method Detail

specialEndKey

private static <R extends KeyRange> byte[] specialEndKey(R range)

SPECIAL CASE wrapper for empty end key

Returns:

ENDKEY if end key is empty, else normal endkey.

add

public boolean add(R range)

Adds an edge to the split calculator

Returns:

true if is included, false if backwards/invalid

calcCoverage

public org.apache.hbase.thirdparty.com.google.common.collect.Multimap<byte[],R> calcCoverage()

Generates a coverage multimap from split key to Regions that start with the split key.

Returns:

coverage multimap

getSplits

public TreeSet<byte[]> getSplits()

getStarts

public org.apache.hbase.thirdparty.com.google.common.collect.Multimap<byte[],R> getStarts()

findBigRanges

public static <R extends KeyRange> List<R> findBigRanges(Collection<R> bigOverlap,
                                                         int count)

Find specified number of top ranges in a big overlap group. It could return less if there are not that many top ranges. Once these top ranges are excluded, the big overlap group will be broken into ranges with no overlapping, or smaller overlapped groups, and most likely some holes.

Parameters:

bigOverlap - a list of ranges that overlap with each other

count - the max number of ranges to find

Returns:

a list of ranges that overlap with most others