org.apache.drill.exec.store.easy.text.reader

Class BaseFieldOutput

java.lang.Object

org.apache.drill.exec.store.easy.text.reader.BaseFieldOutput

Direct Known Subclasses:

ConstrainedFieldOutput, RepeatedVarCharOutput

public abstract class BaseFieldOutput
extends Object

Field Summary

Fields

Modifier and Type	Field and Description
protected int	currentDataPointer
protected int	currentFieldIndex
protected byte[]	fieldBytes
protected boolean	fieldProjected
protected int	fieldWriteCount
protected int	maxField
protected RowSetLoader	writer

Constructor Summary

Constructors

Constructor and Description
BaseFieldOutput(RowSetLoader writer, int maxField, boolean[] projectionMask) Initialize the field output for one of three scenarios: SELECT all: SELECT *, SELECT columns.

Method Summary

Modifier and Type	Method and Description
void	append(byte data) Appends a byte to the output character data buffer.
protected abstract ValueWriter	columnWriter()
boolean	endEmptyField() Shortcut that lets the output know that we are closing ending a field with no data.
boolean	endField() End processing a field within a record.
void	finishRecord() Completes the processing of a given record.
long	getRecordCount() Return the total number of records (across batches) processed
boolean	isFull() Indicates if the current batch is full and reading for this batch should stop.
void	startField(int index) Start processing a new field within a record.
void	startRecord() Start a new record record.
protected void	writeToVector() Write a buffer of data to the underlying vector using the column writer.

Methods inherited from class java.lang.Object

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

Field Detail

currentFieldIndex

protected int currentFieldIndex

currentDataPointer

protected int currentDataPointer

fieldWriteCount

protected int fieldWriteCount

fieldBytes

protected byte[] fieldBytes

writer

protected final RowSetLoader writer

maxField

protected final int maxField

fieldProjected

protected boolean fieldProjected

Constructor Detail

BaseFieldOutput

public BaseFieldOutput(RowSetLoader writer,
                       int maxField,
                       boolean[] projectionMask)

Initialize the field output for one of three scenarios:

• SELECT all: SELECT *, SELECT columns. Indicated by a non -1 max fields.
• SELECT none: SELECT COUNT(*), etc. Indicated by a max field of -1.
• SELECT a, b, c indicated by a non-null projection mask that identifies the indexes of the fields to be selected. In this case, this constructor computes the maximum field.

Parameters:

writer - Row set writer that provides access to the writer for each column

maxField - the index of the last field to store. May be -1 if no fields are to be stored. Computed if the projection mask is set

projectionMask - a boolean array indicating which fields are to be projected to the output. Optional

Method Detail

startRecord

public void startRecord()

Start a new record record. Resets all pointers

startField

public void startField(int index)

Start processing a new field within a record.

Parameters:

index - index within the record

append

public void append(byte data)

Appends a byte to the output character data buffer.

Parameters:

data - current byte read

writeToVector

protected void writeToVector()

Write a buffer of data to the underlying vector using the column writer. The buffer holds a complete or partial chunk of data for the field. If this is the first data for the field, write the bytes. If this is a second buffer for the same field, append the bytes. The append will work if the underlying vector is VarChar, it will fail if a type conversion shim is in between. (This is generally OK because the previous setBytes should have failed because a large int or date is not supported.)

columnWriter

protected abstract ValueWriter columnWriter()

endField

public boolean endField()

End processing a field within a record.

Returns:

true if engine should continue processing record. false if rest of record can be skipped.

endEmptyField

public boolean endEmptyField()

Shortcut that lets the output know that we are closing ending a field with no data.

Returns:

true if engine should continue processing record. false if rest of record can be skipped.

finishRecord

public void finishRecord()

Completes the processing of a given record. Also completes the processing of the last field being read.

getRecordCount

public long getRecordCount()

Return the total number of records (across batches) processed

Returns:

record count

isFull

public boolean isFull()

Indicates if the current batch is full and reading for this batch should stop.

Returns:

true if the batch is full and the reader must exit to allow the batch to be sent downstream, false if the reader may continue to add rows to the current batch