org.janusgraph.diskstorage.keycolumnvalue.scan

Interface ScanJob

All Superinterfaces:

Cloneable

All Known Implementing Classes:

AbstractScanJob, GhostVertexRemover, IndexRemoveJob, VertexJobConverter, VertexMapJob.Executor, VertexProgramScanJob.Executor

public interface ScanJob
extends Cloneable

A global computation over

Author:

Matthias Broecheler ([email protected])

Method Summary

Modifier and Type	Method and Description
ScanJob	clone() Returns a clone of this ScanJob.
default Predicate<StaticBuffer>	getKeyFilter() A predicate that determines whether process(org.janusgraph.diskstorage.StaticBuffer, java.util.Map, ScanMetrics) should be invoked for the given key.
List<SliceQuery>	getQueries() Returns one or more SliceQuery instances belonging to this ScanJob.
void	process(StaticBuffer key, Map<SliceQuery,EntryList> entries, ScanMetrics metrics) Run this ScanJob's computation on the supplied row-key and entries.
default void	workerIterationEnd(ScanMetrics metrics) Invoked after a block of computation (i.e.
default void	workerIterationStart(Configuration jobConfiguration, Configuration graphConfiguration, ScanMetrics metrics) Invoked before a block of computation (i.e.

Method Detail

workerIterationStart

default void workerIterationStart(Configuration jobConfiguration,
                                  Configuration graphConfiguration,
                                  ScanMetrics metrics)

Invoked before a block of computation (i.e. multiple process() calls) is handed to this particular ScanJob. Can be used to initialize the iteration. This method is called exactly once for each before a block of computation. This method is semantically aligned with VertexProgram.workerIterationStart(Memory) This method may not be called if there is no data to be processed. Correspondingly, the end method won't be called either. No-op default implementation.

Parameters:

jobConfiguration - configuration for this particular job

graphConfiguration - configuration options for the entire graph against which this job is executed

metrics - ScanMetrics for this job

workerIterationEnd

default void workerIterationEnd(ScanMetrics metrics)

Invoked after a block of computation (i.e. multiple process() calls) is handed to this particular ScanJob. Can be used to close any resources held by this job. This method is called exactly once for each after a block of computation. This method is semantically aligned with VertexProgram.workerIterationEnd(Memory) This method may not be called if there is no data to be processed. Correspondingly, the start method won't be called either. No-op default implementation.

Parameters:

metrics - ScanMetrics for this job

process

void process(StaticBuffer key,
             Map<SliceQuery,EntryList> entries,
             ScanMetrics metrics)

Run this ScanJob's computation on the supplied row-key and entries.

This method will be called by a client of this interface if and only if both of the following criteria are satisfied:

• The predicate returned by getKeyFilter() must evaluate to true on the key.
• The entries parameter must contain an entry whose key is the first SliceQuery returned by getQueries() and whose value is an EntryList with at least size one. In other words, the initial query in this job's query list must have matched at least once.

Implementations may assume these two conditions are satisfied. Calling this method when either of these two conditions is false yields undefined behavior.

It is the caller's responsibility to construct an entries map that maps each SliceQuery to that query's matches. The caller is also responsible for truncating the entries values to honor BaseQuery.getLimit() when BaseQuery.hasLimit() is true. Passing in an entries value longer than the limit set in its respective key yields undefined behavior.

This method may be called by concurrent threads in a single process.

Parameters:

key -

entries -

metrics -

getQueries

List<SliceQuery> getQueries()

Returns one or more SliceQuery instances belonging to this ScanJob.

Before calling process(org.janusgraph.diskstorage.StaticBuffer, java.util.Map, ScanMetrics), users of this interface must check that the key in question contains at least one entry matching the initial SliceQuery returned by this method. See the javadoc for the process method for more information.

If this method returns more than one query, then the initial query's lower bound must be all zero bits and the initial query's upper bound must be all one bits (per the preconditions in StandardScannerExecutor, the reference ScanJob executor).

Returns:

one or more queries

getKeyFilter

default Predicate<StaticBuffer> getKeyFilter()

A predicate that determines whether process(org.janusgraph.diskstorage.StaticBuffer, java.util.Map, ScanMetrics) should be invoked for the given key. If the predicate returns true, then users of this interface should invoke process for the key and its associated entries. If the predicate returns false, then users of this interface need not invoke process for the key and its associated entries.

This is essentially an optimization that lets implementations of this interface signal to client code that a row can be safely skipped without affecting the execution of this ScanJob.

The returned predicate may be called by concurrent threads in a single process.

Returns:

a threadsafe predicate for edgestore keys

clone

ScanJob clone()

Returns a clone of this ScanJob. The clone will not yet be initialized for computation but all of its internal state (if any) must match that of the original copy.

Returns:

A clone of this ScanJob