ml.shifu.guagua.worker

Class AbstractCombineWorkerComputable<MASTER_RESULT extends Bytable,WORKER_RESULT extends Bytable,KEY extends Bytable,VALUE extends Bytable>

java.lang.Object

ml.shifu.guagua.worker.AbstractCombineWorkerComputable<MASTER_RESULT,WORKER_RESULT,KEY,VALUE>

Type Parameters:

MASTER_RESULT - master result for computation in each iteration.

WORKER_RESULT - worker result for computation in each iteration.

KEY - key type for each record

VALUE - value type for each record

All Implemented Interfaces:

WorkerComputable<MASTER_RESULT,WORKER_RESULT>

public abstract class AbstractCombineWorkerComputable<MASTER_RESULT extends Bytable,WORKER_RESULT extends Bytable,KEY extends Bytable,VALUE extends Bytable>
extends Object
implements WorkerComputable<MASTER_RESULT,WORKER_RESULT>

A high-effective implementation to load data and do computation. This is different with AbstractWorkerComputable, only doCompute(Bytable, Bytable, WorkerContext) for each record are published to user. But the first iteration to load data is included in computation.

Worker result should be updated in doCompute(Bytable, Bytable, WorkerContext), and which will also be populated to Master when all records are processed in one iteration.

To load data successfully, make sure GuaguaRecordReader is initialized firstly. in initRecordReader(GuaguaFileSplit):

 this.setRecordReader(new GuaguaSequenceAsTextRecordReader());
 this.getRecordReader().initialize(fileSplit);
 

or directly use other constructors:

 this.setRecordReader(new GuaguaSequenceAsTextRecordReader(fileSplit));
 

After data is loaded in the first iteration, one can store the data into collections (meomory or disk) to do later iteration logic. But OOM issue should be taken care by users.

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	AbstractCombineWorkerComputable()
protected	AbstractCombineWorkerComputable(boolean isOrder)

Method Summary

Modifier and Type	Method and Description
WORKER_RESULT	compute(WorkerContext<MASTER_RESULT,WORKER_RESULT> context) Worker computation for each iteration.
abstract void	doCompute(KEY currentKey, VALUE currentValue, WorkerContext<MASTER_RESULT,WORKER_RESULT> context) Computation by each record, all update can be set to WORKER_RESULT by context.setCurrentWorkerResult(WORKER_RESULT);
GuaguaRecordReader<KEY,VALUE>	getRecordReader()
abstract void	init(WorkerContext<MASTER_RESULT,WORKER_RESULT> workerContext) Initialization work for the whole computation
abstract void	initRecordReader(GuaguaFileSplit fileSplit) Each GuaguaFileSplit must be initialized before loading data.
protected void	postLoad(WorkerContext<MASTER_RESULT,WORKER_RESULT> workerContext) Do some post work after loading data.
protected void	preLoad(WorkerContext<MASTER_RESULT,WORKER_RESULT> workerContext) Do some pre work before loading data.
void	setRecordReader(GuaguaRecordReader<KEY,VALUE> recordReader)

Methods inherited from class java.lang.Object

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

Constructor Detail

AbstractCombineWorkerComputable

protected AbstractCombineWorkerComputable()

AbstractCombineWorkerComputable

protected AbstractCombineWorkerComputable(boolean isOrder)

Method Detail

compute

public WORKER_RESULT compute(WorkerContext<MASTER_RESULT,WORKER_RESULT> context)
                      throws IOException

Description copied from interface: WorkerComputable

Worker computation for each iteration.

Specified by:

compute in interface WorkerComputable<MASTER_RESULT extends Bytable,WORKER_RESULT extends Bytable>

Parameters:

context - the worker context instance which includes worker info, master result of last iteration or other useful into for each iteration.

Returns:

the worker result of each iteration.

Throws:

IOException - if any io exception in computation, for example, IOException in reading data.

preLoad

protected void preLoad(WorkerContext<MASTER_RESULT,WORKER_RESULT> workerContext)

Do some pre work before loading data.

postLoad

protected void postLoad(WorkerContext<MASTER_RESULT,WORKER_RESULT> workerContext)

Do some post work after loading data.

initRecordReader

public abstract void initRecordReader(GuaguaFileSplit fileSplit)
                               throws IOException

Each GuaguaFileSplit must be initialized before loading data.

Throws:

IOException

init

public abstract void init(WorkerContext<MASTER_RESULT,WORKER_RESULT> workerContext)

Initialization work for the whole computation

doCompute

public abstract void doCompute(KEY currentKey,
                               VALUE currentValue,
                               WorkerContext<MASTER_RESULT,WORKER_RESULT> context)

Computation by each record, all update can be set to WORKER_RESULT by context.setCurrentWorkerResult(WORKER_RESULT);

getRecordReader

public GuaguaRecordReader<KEY,VALUE> getRecordReader()

setRecordReader

public void setRecordReader(GuaguaRecordReader<KEY,VALUE> recordReader)