Package org.apache.hadoop.hbase.io.hfile

Interface BlockCompressedSizePredicator

All Known Implementing Classes:

PreviousBlockCompressionRatePredicator, UncompressedBlockSizePredicator

@Private
public interface BlockCompressedSizePredicator

Allows for defining different compression rate predicates on its implementing classes. Useful when compression is in place, and we want to define block size based on the compressed size, rather than the default behaviour that considers the uncompressed size only. Since we don't actually know the compressed size until we actual apply compression in the block byte buffer, we need to "predicate" this compression rate and minimize compression execution to avoid excessive resources usage. Different approaches for predicating the compressed block size can be defined by implementing classes. The updateLatestBlockSizes allows for updating uncompressed and compressed size values, and is called during block finishing (when we finally apply compression on the block data). Final block size predicate logic is implemented in shouldFinishBlock, which is called by the block writer once uncompressed size has reached the configured BLOCK size, and additional checks should be applied to decide if the block can be finished.

Field Summary

Fields

Modifier and Type	Field	Description
static final String	BLOCK_COMPRESSED_SIZE_PREDICATOR
static final String	MAX_BLOCK_SIZE_UNCOMPRESSED

Method Summary

Modifier and Type	Method	Description
boolean	shouldFinishBlock(int uncompressed)	Decides if the block should be finished based on the comparison of its uncompressed size against an adjusted size based on a predicated compression factor.
void	updateLatestBlockSizes(HFileContext context, int uncompressed, int compressed)	Updates the predicator with both compressed and uncompressed sizes of latest block written.

Field Details

BLOCK_COMPRESSED_SIZE_PREDICATOR

static final String BLOCK_COMPRESSED_SIZE_PREDICATOR

See Also:

• Constant Field Values

MAX_BLOCK_SIZE_UNCOMPRESSED

static final String MAX_BLOCK_SIZE_UNCOMPRESSED

See Also:

• Constant Field Values

Method Details

updateLatestBlockSizes

void updateLatestBlockSizes(HFileContext context,
 int uncompressed,
 int compressed)

Updates the predicator with both compressed and uncompressed sizes of latest block written. To be called once the block is finshed and flushed to disk after compression.

Parameters:

context - the HFileContext containg the configured max block size.

uncompressed - the uncompressed size of last block written.

compressed - the compressed size of last block written.

shouldFinishBlock

boolean shouldFinishBlock(int uncompressed)

Decides if the block should be finished based on the comparison of its uncompressed size against an adjusted size based on a predicated compression factor.

Parameters:

uncompressed - true if the block should be finished.