org.apache.cassandra.service.pager

Class MultiPartitionPager<T extends SinglePartitionReadQuery>

java.lang.Object

org.apache.cassandra.service.pager.MultiPartitionPager<T>

All Implemented Interfaces:

QueryPager

public class MultiPartitionPager<T extends SinglePartitionReadQuery>
extends java.lang.Object
implements QueryPager

Pager over a list of SinglePartitionReadQuery. Note that this is not easy to make efficient. Indeed, we need to page the first query fully before returning results from the next one, but if the result returned by each query is small (compared to pageSize), paging the queries one at a time under-performs compared to parallelizing. On the other, if we parallelize and each query raised pageSize results, we'll end up with queries.size() * pageSize results in memory, which defeats the purpose of paging. For now, we keep it simple (somewhat) and just do one query at a time. Provided that we make sure to not create a pager unless we need to, this is probably fine. Though if we later want to get fancy, we could use the cfs meanPartitionSize to decide if parallelizing some of the query might be worth it while being confident we don't blow out memory.

Field Summary

Fields inherited from interface org.apache.cassandra.service.pager.QueryPager

EMPTY

Constructor Summary

Constructors

Constructor and Description
MultiPartitionPager(SinglePartitionReadQuery.Group<T> group, PagingState state, ProtocolVersion protocolVersion)

Method Summary

Modifier and Type	Method and Description
ReadExecutionController	executionController() Starts a new read operation.
PartitionIterator	fetchPage(int pageSize, ConsistencyLevel consistency, ClientState clientState, long queryStartNanoTime) Fetches the next page.
PartitionIterator	fetchPageInternal(int pageSize, ReadExecutionController executionController) Fetches the next page internally (in other, this does a local query).
boolean	isExhausted() Whether or not this pager is exhausted, i.e.
int	maxRemaining() The maximum number of cells/CQL3 row that we may still have to return.
PagingState	state() Get the current state of the pager.
QueryPager	withUpdatedLimit(DataLimits newLimits) Creates a new QueryPager that use the new limits.

Methods inherited from class java.lang.Object

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

Constructor Detail

MultiPartitionPager

public MultiPartitionPager(SinglePartitionReadQuery.Group<T> group,
                           PagingState state,
                           ProtocolVersion protocolVersion)

Method Detail

withUpdatedLimit

public QueryPager withUpdatedLimit(DataLimits newLimits)

Description copied from interface: QueryPager

Creates a new QueryPager that use the new limits.

Specified by:

withUpdatedLimit in interface QueryPager

Parameters:

newLimits - the new limits

Returns:

a new QueryPager that use the new limits

state

public PagingState state()

Description copied from interface: QueryPager

Get the current state of the pager. The state can allow to restart the paging on another host from where we are at this point.

Specified by:

state in interface QueryPager

Returns:

the current paging state. Will return null if paging is at the beginning. If the pager is exhausted, the result is undefined.

isExhausted

public boolean isExhausted()

Description copied from interface: QueryPager

Whether or not this pager is exhausted, i.e. whether or not a call to fetchPage may return more result.

Specified by:

isExhausted in interface QueryPager

Returns:

whether the pager is exhausted.

executionController

public ReadExecutionController executionController()

Description copied from interface: QueryPager

Starts a new read operation.

This must be called before fetchPageInternal and passed to it to protect the read. The returned object must be closed on all path and it is thus strongly advised to use it in a try-with-ressource construction.

Specified by:

executionController in interface QueryPager

Returns:

a newly started order group for this QueryPager.

fetchPage

public PartitionIterator fetchPage(int pageSize,
                                   ConsistencyLevel consistency,
                                   ClientState clientState,
                                   long queryStartNanoTime)
                            throws RequestValidationException,
                                   RequestExecutionException

Description copied from interface: QueryPager

Fetches the next page.

Specified by:

fetchPage in interface QueryPager

Parameters:

pageSize - the maximum number of elements to return in the next page.

consistency - the consistency level to achieve for the query.

clientState - the ClientState for the query. In practice, this can be null unless consistency is a serial consistency.

Returns:

the page of result.

Throws:

RequestValidationException

RequestExecutionException

fetchPageInternal

public PartitionIterator fetchPageInternal(int pageSize,
                                           ReadExecutionController executionController)
                                    throws RequestValidationException,
                                           RequestExecutionException

Description copied from interface: QueryPager

Fetches the next page internally (in other, this does a local query).

Specified by:

fetchPageInternal in interface QueryPager

Parameters:

pageSize - the maximum number of elements to return in the next page.

executionController - the ReadExecutionController protecting the read.

Returns:

the page of result.

Throws:

RequestValidationException

RequestExecutionException

maxRemaining

public int maxRemaining()

Description copied from interface: QueryPager

The maximum number of cells/CQL3 row that we may still have to return. In other words, that's the initial user limit minus what we've already returned (note that it's not how many we *will* return, just the upper limit on it).

Specified by:

maxRemaining in interface QueryPager