org.apache.hadoop.hbase.codec.prefixtree.encode.column

Class ColumnSectionWriter

java.lang.Object

org.apache.hadoop.hbase.codec.prefixtree.encode.column.ColumnSectionWriter

@InterfaceAudience.Private
public class ColumnSectionWriter
extends Object

Takes the tokenized family or qualifier data and flattens it into a stream of bytes. The family section is written after the row section, and qualifier section after family section.

The family and qualifier tries, or "column tries", are structured differently than the row trie. The trie cannot be reassembled without external data about the offsets of the leaf nodes, and these external pointers are stored in the nubs and leaves of the row trie. For each cell in a row, the row trie contains a list of offsets into the column sections (along with pointers to timestamps and other per-cell fields). These offsets point to the last column node/token that comprises the column name. To assemble the column name, the trie is traversed in reverse (right to left), with the rightmost tokens pointing to the start of their "parent" node which is the node to the left.

This choice was made to reduce the size of the column trie by storing the minimum amount of offset data. As a result, to find a specific qualifier within a row, you must do a binary search of the column nodes, reassembling each one as you search. Future versions of the PrefixTree might encode the columns in both a forward and reverse trie, which would convert binary searches into more efficient trie searches which would be beneficial for wide rows.

Field Summary

Fields

Modifier and Type	Field and Description
static int	EXPECTED_NUBS_PLUS_LEAVES

Constructor Summary

Constructors

Constructor and Description
ColumnSectionWriter() construct
ColumnSectionWriter(PrefixTreeBlockMeta blockMeta, Tokenizer builder, ColumnNodeType nodeType)

Method Summary

Methods

Modifier and Type	Method and Description
ColumnSectionWriter	compile() methods
protected void	compilerInternals()
ArrayList<ColumnNodeWriter>	getColumnNodeWriters() get/set
ArrayList<TokenizerNode>	getLeaves()
ArrayList<TokenizerNode>	getNonLeaves()
int	getNumBytes()
int	getOutputArrayOffset(int sortedIndex)
void	reconstruct(PrefixTreeBlockMeta blockMeta, Tokenizer builder, ColumnNodeType nodeType)
void	reset()
void	writeBytes(OutputStream os)

Methods inherited from class java.lang.Object

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

Field Detail

EXPECTED_NUBS_PLUS_LEAVES

public static final int EXPECTED_NUBS_PLUS_LEAVES

See Also:

Constant Field Values

Constructor Detail

ColumnSectionWriter

public ColumnSectionWriter()

construct

ColumnSectionWriter

public ColumnSectionWriter(PrefixTreeBlockMeta blockMeta,
                   Tokenizer builder,
                   ColumnNodeType nodeType)

Method Detail

reconstruct

public void reconstruct(PrefixTreeBlockMeta blockMeta,
               Tokenizer builder,
               ColumnNodeType nodeType)

reset

public void reset()

compile

public ColumnSectionWriter compile()

methods

compilerInternals

protected void compilerInternals()

writeBytes

public void writeBytes(OutputStream os)
                throws IOException

Throws:

IOException

getColumnNodeWriters

public ArrayList<ColumnNodeWriter> getColumnNodeWriters()

get/set

getNumBytes

public int getNumBytes()

getOutputArrayOffset

public int getOutputArrayOffset(int sortedIndex)

getNonLeaves

public ArrayList<TokenizerNode> getNonLeaves()

getLeaves

public ArrayList<TokenizerNode> getLeaves()