org.apache.cassandra.locator

Class RangesAtEndpoint

java.lang.Object

org.apache.cassandra.locator.AbstractReplicaCollection<RangesAtEndpoint>

org.apache.cassandra.locator.RangesAtEndpoint

All Implemented Interfaces:

java.lang.Iterable<Replica>, ReplicaCollection<RangesAtEndpoint>

Direct Known Subclasses:

RangesAtEndpoint.Builder

public class RangesAtEndpoint
extends AbstractReplicaCollection<RangesAtEndpoint>

A ReplicaCollection for Ranges occurring at an endpoint. All Replica will be for the same endpoint, and must be unique Ranges (though overlapping ranges are presently permitted, these should probably not be permitted to occur)

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	RangesAtEndpoint.Builder

Nested classes/interfaces inherited from class org.apache.cassandra.locator.AbstractReplicaCollection

AbstractReplicaCollection.ReplicaList, AbstractReplicaCollection.ReplicaMap<K>

Field Summary

Fields inherited from class org.apache.cassandra.locator.AbstractReplicaCollection

EMPTY_LIST, list

Method Summary

Modifier and Type	Method and Description
static RangesAtEndpoint.Builder	builder(InetAddressAndPort endpoint)
static RangesAtEndpoint.Builder	builder(InetAddressAndPort endpoint, int capacity)
java.util.Map<Range<Token>,Replica>	byRange()
static java.util.stream.Collector<Replica,RangesAtEndpoint.Builder,RangesAtEndpoint>	collector(InetAddressAndPort endpoint)
static RangesAtEndpoint	concat(RangesAtEndpoint replicas, RangesAtEndpoint extraReplicas)
boolean	contains(Range<Token> range, boolean isFull)
boolean	contains(Replica replica)
static RangesAtEndpoint	copyOf(java.util.List<Replica> replicas)
static RangesAtEndpoint	empty(InetAddressAndPort endpoint)
InetAddressAndPort	endpoint()
java.util.Set<InetAddressAndPort>	endpoints()
static boolean	isDummyList(RangesAtEndpoint ranges)
ReplicaCollection.Builder<RangesAtEndpoint>	newBuilder(int initialCapacity) construct a new Builder of our own type, so that we can concatenate TODO: this isn't terribly pretty, but we need sometimes to select / merge two Endpoints of unknown type;
static RangesAtEndpoint	of(Replica... replicas)
static RangesAtEndpoint	of(Replica replica)
RangesAtEndpoint	onlyFull()
RangesAtEndpoint	onlyTransient()
java.util.Set<Range<Token>>	ranges()
RangesAtEndpoint	snapshot()
protected RangesAtEndpoint	snapshot(AbstractReplicaCollection.ReplicaList newList)
static RangesAtEndpoint	toDummyList(java.util.Collection<Range<Token>> ranges) Use of this method to synthesize Replicas is almost always wrong.
RangesAtEndpoint	unwrap()

Methods inherited from class org.apache.cassandra.locator.AbstractReplicaCollection

collector, count, equals, filter, filter, filterLazily, filterLazily, get, hashCode, isEmpty, iterator, size, sorted, stream, subList, toString

Methods inherited from class java.lang.Object

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

Methods inherited from interface java.lang.Iterable

forEach, spliterator

Method Detail

endpoint

public InetAddressAndPort endpoint()

endpoints

public java.util.Set<InetAddressAndPort> endpoints()

Returns:

a Set of the endpoints of the contained Replicas. Iteration order is maintained where there is a 1:1 relationship between endpoint and Replica Typically this collection offers O(1) access methods, and this is true for all but ReplicaList.

ranges

public java.util.Set<Range<Token>> ranges()

Returns:

a set of all unique Ranges This method is threadsafe, though it is not synchronised

byRange

public java.util.Map<Range<Token>,Replica> byRange()

Returns:

a map of all Ranges, to their owning Replica instance This method is threadsafe, though it is not synchronised

snapshot

protected RangesAtEndpoint snapshot(AbstractReplicaCollection.ReplicaList newList)

snapshot

public RangesAtEndpoint snapshot()

Specified by:

snapshot in class AbstractReplicaCollection<RangesAtEndpoint>

newBuilder

public ReplicaCollection.Builder<RangesAtEndpoint> newBuilder(int initialCapacity)

Description copied from class: AbstractReplicaCollection

construct a new Builder of our own type, so that we can concatenate TODO: this isn't terribly pretty, but we need sometimes to select / merge two Endpoints of unknown type;

Specified by:

newBuilder in class AbstractReplicaCollection<RangesAtEndpoint>

contains

public boolean contains(Replica replica)

Returns:

true iff a Replica in this collection is equal to the provided Replica. Typically this method is expected to take O(1) time, and this is true for all but ReplicaList.

onlyFull

public RangesAtEndpoint onlyFull()

onlyTransient

public RangesAtEndpoint onlyTransient()

contains

public boolean contains(Range<Token> range,
                        boolean isFull)

unwrap

public RangesAtEndpoint unwrap()

Returns:

if there are no wrap around ranges contained in this RangesAtEndpoint, return self; otherwise, return a RangesAtEndpoint covering the same logical portions of the ring, but with those ranges unwrapped

collector

public static java.util.stream.Collector<Replica,RangesAtEndpoint.Builder,RangesAtEndpoint> collector(InetAddressAndPort endpoint)

builder

public static RangesAtEndpoint.Builder builder(InetAddressAndPort endpoint)

builder

public static RangesAtEndpoint.Builder builder(InetAddressAndPort endpoint,
                                               int capacity)

empty

public static RangesAtEndpoint empty(InetAddressAndPort endpoint)

of

public static RangesAtEndpoint of(Replica replica)

of

public static RangesAtEndpoint of(Replica... replicas)

copyOf

public static RangesAtEndpoint copyOf(java.util.List<Replica> replicas)

toDummyList

public static RangesAtEndpoint toDummyList(java.util.Collection<Range<Token>> ranges)

Use of this method to synthesize Replicas is almost always wrong. In repair it turns out the concerns of transient vs non-transient are handled at a higher level, but eventually repair needs to ask streaming to actually move the data and at that point it doesn't have a great handle on what the replicas are and it doesn't really matter. Streaming expects to be given Replicas with each replica indicating what type of data (transient or not transient) should be sent. So in this one instance we can lie to streaming and pretend all the replicas are full and use a dummy address and it doesn't matter because streaming doesn't rely on the address for anything other than debugging and full is a valid value for transientness because streaming is selecting candidate tables from the repair/unrepaired set already.

Parameters:

ranges -

Returns:

isDummyList

public static boolean isDummyList(RangesAtEndpoint ranges)

concat

public static RangesAtEndpoint concat(RangesAtEndpoint replicas,
                                      RangesAtEndpoint extraReplicas)

Returns:

concatenate two DISJOINT collections together