parquet.hadoop.api

Class WriteSupport<T>

java.lang.Object

parquet.hadoop.api.WriteSupport<T>

Type Parameters:

T - the type of the incoming records

Direct Known Subclasses:

DelegatingWriteSupport, GroupWriteSupport

public abstract class WriteSupport<T>
extends Object

Abstraction to use with ParquetOutputFormat to convert incoming records

Author:

Julien Le Dem

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	WriteSupport.FinalizedWriteContext Information to be added in the file once all the records have been written
static class	WriteSupport.WriteContext information to be persisted in the file

Constructor Summary

Constructors

Constructor and Description
WriteSupport()

Method Summary

Methods

Modifier and Type	Method and Description
WriteSupport.FinalizedWriteContext	finalizeWrite() called once in the end after the last record was written
abstract WriteSupport.WriteContext	init(org.apache.hadoop.conf.Configuration configuration) called first in the task
abstract void	prepareForWrite(RecordConsumer recordConsumer) This will be called once per row group
abstract void	write(T record) called once per record

Methods inherited from class java.lang.Object

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

Constructor Detail

WriteSupport

public WriteSupport()

Method Detail

init

public abstract WriteSupport.WriteContext init(org.apache.hadoop.conf.Configuration configuration)

called first in the task

Parameters:

configuration - the job's configuration

Returns:

the information needed to write the file

prepareForWrite

public abstract void prepareForWrite(RecordConsumer recordConsumer)

This will be called once per row group

Parameters:

recordConsumer - the recordConsumer to write to

write

public abstract void write(T record)

called once per record

Parameters:

record - one record to write to the previously provided record consumer

finalizeWrite

public WriteSupport.FinalizedWriteContext finalizeWrite()

called once in the end after the last record was written

Returns:

information to be added in the file