Package org.apache.drill.exec.store.parquet.columnreaders.batchsizing

Class RecordBatchSizerManager

java.lang.Object

org.apache.drill.exec.store.parquet.columnreaders.batchsizing.RecordBatchSizerManager

public final class RecordBatchSizerManager
extends Object

This class is tasked with managing all aspects of flat Parquet reader record batch sizing logic. Currently a record batch size is constrained with two parameters: Number of rows and Memory usage.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static final class	RecordBatchSizerManager.ColumnMemoryQuota	Field memory quota
static interface	RecordBatchSizerManager.FieldOverflowState	An abstraction to allow column readers attach custom field overflow state
static final class	RecordBatchSizerManager.FieldOverflowStateContainer	Container object to hold current field overflow state
static final class	RecordBatchSizerManager.VarLenColumnBatchStats	Container object to supply variable columns statistics to the batch sizer

Constructor Summary

Constructors

Constructor	Description
RecordBatchSizerManager(OptionManager options, ParquetSchema schema, long totalRecordsToRead, RecordBatchStats.RecordBatchStatsContext batchStatsContext)	Constructor.

Method Summary

Modifier and Type	Method	Description
void	allocate(Map<String,ValueVector> vectorMap)	Allocates value vectors for the current batch.
void	close()	Closes all resources managed by this object
RecordBatchStats.RecordBatchStatsContext	getBatchStatsContext()
long	getConfigMemorySizePerBatch()
int	getConfigRecordsPerBatch()
RecordBatchSizerManager.ColumnMemoryQuota	getCurrentFieldBatchMemory(String field)
long	getCurrentMemorySizePerBatch()
int	getCurrentRecordsPerBatch()
RecordBatchSizerManager.FieldOverflowStateContainer	getFieldOverflowContainer(String field)
Map<String,RecordBatchSizerManager.FieldOverflowStateContainer>	getFieldOverflowMap()
ParquetSchema	getSchema()
void	onEndOfBatch(int batchNumRecords, List<RecordBatchSizerManager.VarLenColumnBatchStats> batchStats)	Enables this object to optimize the impact of overflows by computing more accurate VL column precision.
boolean	releaseFieldOverflowContainer(String field)	Releases the overflow data resources associated with this field; also removes the overflow container from the overflow containers map.
void	setup()	Tunes record batch parameters based on configuration and schema.

Methods inherited from class java.lang.Object

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

Constructor Details

RecordBatchSizerManager

public RecordBatchSizerManager(OptionManager options,
 ParquetSchema schema,
 long totalRecordsToRead,
 RecordBatchStats.RecordBatchStatsContext batchStatsContext)

Constructor.

Parameters:

options - drill options

schema - current reader schema

totalRecordsToRead - total number of rows to read

Method Details

setup

public void setup()

Tunes record batch parameters based on configuration and schema.

getSchema

public ParquetSchema getSchema()

Returns:

the schema

getBatchStatsContext

public RecordBatchStats.RecordBatchStatsContext getBatchStatsContext()

Returns:

batch statistics context

allocate

public void allocate(Map<String,ValueVector> vectorMap)
              throws OutOfMemoryException

Allocates value vectors for the current batch.

Parameters:

vectorMap - a collection of value vectors keyed by their field names

Throws:

OutOfMemoryException

getFieldOverflowMap

public Map<String,RecordBatchSizerManager.FieldOverflowStateContainer> getFieldOverflowMap()

Returns:

the field overflow state map

getFieldOverflowContainer

public RecordBatchSizerManager.FieldOverflowStateContainer getFieldOverflowContainer(String field)

Parameters:

field - materialized field

Returns:

field overflow state container

releaseFieldOverflowContainer

public boolean releaseFieldOverflowContainer(String field)

Releases the overflow data resources associated with this field; also removes the overflow container from the overflow containers map.

Parameters:

field - materialized field

Returns:

true if this field's overflow container was removed from the overflow containers map

getCurrentFieldBatchMemory

public RecordBatchSizerManager.ColumnMemoryQuota getCurrentFieldBatchMemory(String field)

Parameters:

field - materialized field

Returns:

field batch memory quota

getCurrentRecordsPerBatch

public int getCurrentRecordsPerBatch()

Returns:

current number of records per batch (may change across batches)

getCurrentMemorySizePerBatch

public long getCurrentMemorySizePerBatch()

Returns:

current total memory per batch (may change across batches)

getConfigRecordsPerBatch

public int getConfigRecordsPerBatch()

Returns:

configured number of records per batch (may be different from the enforced one)

getConfigMemorySizePerBatch

public long getConfigMemorySizePerBatch()

Returns:

configured memory size per batch (may be different from the enforced one)

onEndOfBatch

public void onEndOfBatch(int batchNumRecords,
 List<RecordBatchSizerManager.VarLenColumnBatchStats> batchStats)

Enables this object to optimize the impact of overflows by computing more accurate VL column precision.

Parameters:

batchNumRecords - number of records in this batch

batchStats - columns statistics

close

public void close()

Closes all resources managed by this object