Interface Block

All Known Subinterfaces:

ValueBlock

All Known Implementing Classes:

ArrayBlock, ByteArrayBlock, DictionaryBlock, Fixed12Block, Int128ArrayBlock, IntArrayBlock, LazyBlock, LongArrayBlock, MapBlock, RowBlock, RunLengthEncodedBlock, ShortArrayBlock, VariableWidthBlock

public sealed interface Block
permits DictionaryBlock, RunLengthEncodedBlock, LazyBlock, ValueBlock

Method Summary

Modifier and Type	Method	Description
Block	copyPositions(int[] positions, int offset, int length)	Returns a block containing the specified positions.
Block	copyRegion(int position, int length)	Returns a block starting at the specified position and extends for the specified length.
Block	copyWithAppendedNull()	Returns a block that contains a copy of the contents of the current block, and an appended null at the end.
OptionalInt	fixedSizeInBytesPerPosition()	Returns the number of bytes (in terms of getSizeInBytes()) required per position that this block contains, assuming that the number of bytes required is a known static quantity and not dependent on any particular specific position.
String	getEncodingName()	Get the encoding for this block.
long	getEstimatedDataSizeForStats(int position)	Returns the estimated in memory data size for stats of position.
default Block	getLoadedBlock()	Returns a fully loaded block that assures all data is in memory.
int	getPositionCount()	Returns the number of positions in this block.
default Block	getPositions(int[] positions, int offset, int length)	Create a new block from the current block by keeping the same elements only with respect to positions that starts at offset and has length of length.
long	getPositionsSizeInBytes(boolean[] positions, int selectedPositionsCount)	Returns the size of all positions marked true in the positions array.
Block	getRegion(int positionOffset, int length)	Returns a block starting at the specified position and extends for the specified length.
long	getRegionSizeInBytes(int position, int length)	Returns the size of block.getRegion(position, length).
long	getRetainedSizeInBytes()	Returns the retained size of this block in memory, including over-allocations.
ValueBlock	getSingleValueBlock(int position)	Gets the value at the specified position as a single element block.
long	getSizeInBytes()	Returns the size of this block as if it was compacted, ignoring any over-allocations and any unloaded nested blocks.
ValueBlock	getUnderlyingValueBlock()	Returns the underlying value block underlying this block.
int	getUnderlyingValuePosition(int position)	Returns the position in the underlying value block corresponding to the specified position in this block.
default boolean	isLoaded()	Returns true if block data is fully loaded into memory.
boolean	isNull(int position)	Is the specified position null?
default boolean	mayHaveNull()	Is it possible the block may have a null value?
void	retainedBytesForEachPart(ObjLongConsumer<Object> consumer)	consumer visits each of the internal data container and accepts the size for it.

Method Details

getSingleValueBlock

ValueBlock getSingleValueBlock(int position)

Gets the value at the specified position as a single element block. The method must copy the data into a new block.

This method is useful for operators that hold on to a single value without holding on to the entire block.

Throws:

IllegalArgumentException - if this position is not valid

getPositionCount

int getPositionCount()

Returns the number of positions in this block.

getSizeInBytes

long getSizeInBytes()

Returns the size of this block as if it was compacted, ignoring any over-allocations and any unloaded nested blocks. For example, in dictionary blocks, this only counts each dictionary entry once, rather than each time a value is referenced.

getRegionSizeInBytes

long getRegionSizeInBytes(int position,
 int length)

Returns the size of block.getRegion(position, length). The method can be expensive. Do not use it outside an implementation of Block.

fixedSizeInBytesPerPosition

OptionalInt fixedSizeInBytesPerPosition()

Returns the number of bytes (in terms of getSizeInBytes()) required per position that this block contains, assuming that the number of bytes required is a known static quantity and not dependent on any particular specific position. This allows for some complex block wrappings to potentially avoid having to call getPositionsSizeInBytes(boolean[], int) which would require computing the specific positions selected

Returns:

The size in bytes, per position, if this block type does not require specific position information to compute its size

getPositionsSizeInBytes

long getPositionsSizeInBytes(boolean[] positions,
 int selectedPositionsCount)

Returns the size of all positions marked true in the positions array. This is equivalent to multiple calls of block.getRegionSizeInBytes(position, length) where you mark all positions for the regions first. The 'selectedPositionsCount' variable may be used to skip iterating through the positions array in case this is a fixed-width block

getRetainedSizeInBytes

long getRetainedSizeInBytes()

Returns the retained size of this block in memory, including over-allocations. This method is called from the inner most execution loop and must be fast.

getEstimatedDataSizeForStats

long getEstimatedDataSizeForStats(int position)

Returns the estimated in memory data size for stats of position. Do not use it for other purpose.

retainedBytesForEachPart

void retainedBytesForEachPart(ObjLongConsumer<Object> consumer)

consumer visits each of the internal data container and accepts the size for it. This method can be helpful in cases such as memory counting for internal data structure. Also, the method should be non-recursive, only visit the elements at the top level, and specifically should not call retainedBytesForEachPart on nested blocks consumer should be called at least once with the current block and must include the instance size of the current block

getEncodingName

String getEncodingName()

Get the encoding for this block.

getPositions

default Block getPositions(int[] positions,
 int offset,
 int length)

Create a new block from the current block by keeping the same elements only with respect to positions that starts at offset and has length of length. The implementation may return a view over the data in this block or may return a copy, and the implementation is allowed to retain the positions array for use in the view.

copyPositions

Block copyPositions(int[] positions,
 int offset,
 int length)

Returns a block containing the specified positions. Positions to copy are stored in a subarray within positions array that starts at offset and has length of length. All specified positions must be valid for this block.

The returned block must be a compact representation of the original block.

getRegion

Block getRegion(int positionOffset,
 int length)

Returns a block starting at the specified position and extends for the specified length. The specified region must be entirely contained within this block.

The region can be a view over this block. If this block is released the region block may also be released. If the region block is released this block may also be released.

copyRegion

Block copyRegion(int position,
 int length)

Returns a block starting at the specified position and extends for the specified length. The specified region must be entirely contained within this block.

The region returned must be a compact representation of the original block, unless their internal representation will be exactly the same. This method is useful for operators that hold on to a range of values without holding on to the entire block.

mayHaveNull

default boolean mayHaveNull()

Is it possible the block may have a null value? If false, the block cannot contain a null, but if true, the block may or may not have a null.

isNull

boolean isNull(int position)

Is the specified position null?

Throws:

IllegalArgumentException - if this position is not valid. The method may return false without throwing exception when there are no nulls in the block, even if the position is invalid

isLoaded

default boolean isLoaded()

Returns true if block data is fully loaded into memory.

getLoadedBlock

default Block getLoadedBlock()

Returns a fully loaded block that assures all data is in memory. Neither the returned block nor any nested block will be a LazyBlock. The same block will be returned if neither the current block nor any nested blocks are LazyBlock,

This allows streaming data sources to skip sections that are not accessed in a query.

copyWithAppendedNull

Block copyWithAppendedNull()

Returns a block that contains a copy of the contents of the current block, and an appended null at the end. The original block will not be modified. The purpose of this method is to leverage the contents of a block and the structure of the implementation to efficiently produce a copy of the block with a NULL element inserted - so that it can be used as a dictionary. This method is expected to be invoked on completely built Block instances i.e. not on in-progress block builders.

getUnderlyingValueBlock

ValueBlock getUnderlyingValueBlock()

Returns the underlying value block underlying this block.

getUnderlyingValuePosition

int getUnderlyingValuePosition(int position)

Returns the position in the underlying value block corresponding to the specified position in this block.