org.apache.hadoop.hbase.regionserver.querymatcher

Class ExplicitColumnTracker

java.lang.Object

org.apache.hadoop.hbase.regionserver.querymatcher.ExplicitColumnTracker

All Implemented Interfaces:

ColumnTracker, ShipperListener

@InterfaceAudience.Private
public class ExplicitColumnTracker
extends Object
implements ColumnTracker

This class is used for the tracking and enforcement of columns and numbers of versions during the course of a Get or Scan operation, when explicit column qualifiers have been asked for in the query. With a little magic (see ScanQueryMatcher), we can use this matcher for both scans and gets. The main difference is 'next' and 'done' collapse for the scan case (since we see all columns in order), and we only reset between rows.

This class is utilized by ScanQueryMatcher mainly through two methods:

• checkColumn(org.apache.hadoop.hbase.Cell, byte) is called when a Put satisfies all other conditions of the query.
• getNextRowOrNextColumn(org.apache.hadoop.hbase.Cell) is called whenever ScanQueryMatcher believes that the current column should be skipped (by timestamp, filter etc.)

These two methods returns a ScanQueryMatcher.MatchCode to define what action should be taken.

This class is NOT thread-safe as queries are never multi-threaded

Constructor Summary

Constructors

Constructor and Description
ExplicitColumnTracker(NavigableSet<byte[]> columns, int minVersions, int maxVersions, long oldestUnexpiredTS) Default constructor.

Method Summary

Modifier and Type	Method and Description
void	beforeShipped() The action that needs to be performed before Shipper.shipped() is performed
ScanQueryMatcher.MatchCode	checkColumn(Cell cell, byte type) Checks if the column is present in the list of requested columns by returning the match code instance.
ScanQueryMatcher.MatchCode	checkVersions(Cell cell, long timestamp, byte type, boolean ignoreCount) Keeps track of the number of versions for the columns asked for.
boolean	done() Done when there are no more columns to match against.
void	doneWithColumn(Cell cell) This method is used to inform the column tracker that we are done with this column.
org.apache.hadoop.hbase.regionserver.querymatcher.ColumnCount	getColumnHint() Used by matcher and scan/get to get a hint of the next column to seek to after checkColumn() returns SKIP.
ScanQueryMatcher.MatchCode	getNextRowOrNextColumn(Cell cell) Retrieve the MatchCode for the next row or column
boolean	isDone(long timestamp) Give the tracker a chance to declare it's done based on only the timestamp to allow an early out.
void	reset() Resets the Matcher

Methods inherited from class java.lang.Object

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

Constructor Detail

ExplicitColumnTracker

public ExplicitColumnTracker(NavigableSet<byte[]> columns,
                             int minVersions,
                             int maxVersions,
                             long oldestUnexpiredTS)

Default constructor.

Parameters:

columns - columns specified user in query

minVersions - minimum number of versions to keep

maxVersions - maximum versions to return per column

oldestUnexpiredTS - the oldest timestamp we are interested in, based on TTL

Method Detail

done

public boolean done()

Done when there are no more columns to match against.

Specified by:

done in interface ColumnTracker

Returns:

true when done.

getColumnHint

public org.apache.hadoop.hbase.regionserver.querymatcher.ColumnCount getColumnHint()

Description copied from interface: ColumnTracker

Used by matcher and scan/get to get a hint of the next column to seek to after checkColumn() returns SKIP. Returns the next interesting column we want, or NULL there is none (wildcard scanner). Implementations aren't required to return anything useful unless the most recent call was to checkColumn() and the return code was SKIP. This is pretty implementation detail-y, but optimizations are like that.

Specified by:

getColumnHint in interface ColumnTracker

Returns:

null, or a ColumnCount that we should seek to

checkColumn

public ScanQueryMatcher.MatchCode checkColumn(Cell cell,
                                              byte type)

Checks if the column is present in the list of requested columns by returning the match code instance. It does not check against the number of versions for the columns asked for. To do the version check, one has to call ColumnTracker.checkVersions(Cell, long, byte, boolean) method based on the return type (INCLUDE) of this method. The values that can be returned by this method are ScanQueryMatcher.MatchCode.INCLUDE, ScanQueryMatcher.MatchCode.SEEK_NEXT_COL and ScanQueryMatcher.MatchCode.SEEK_NEXT_ROW.

Specified by:

checkColumn in interface ColumnTracker

Parameters:

cell - a cell with the column to match against

type - The type of the Cell

Returns:

The match code instance.

checkVersions

public ScanQueryMatcher.MatchCode checkVersions(Cell cell,
                                                long timestamp,
                                                byte type,
                                                boolean ignoreCount)
                                         throws IOException

Description copied from interface: ColumnTracker

Keeps track of the number of versions for the columns asked for. It assumes that the user has already checked if the cell needs to be included by calling the ColumnTracker.checkColumn(Cell, byte) method. The enum values returned by this method are ScanQueryMatcher.MatchCode.SKIP, ScanQueryMatcher.MatchCode.INCLUDE, ScanQueryMatcher.MatchCode.INCLUDE_AND_SEEK_NEXT_COL and ScanQueryMatcher.MatchCode.INCLUDE_AND_SEEK_NEXT_ROW. Implementations which include all the columns could just return ScanQueryMatcher.MatchCode.INCLUDE in the ColumnTracker.checkColumn(Cell, byte) method and perform all the operations in this checkVersions method.

Specified by:

checkVersions in interface ColumnTracker

Parameters:

cell - a cell with the column to match against

timestamp - The timestamp of the cell.

type - the type of the key value (Put/Delete)

ignoreCount - indicates if the KV needs to be excluded while counting (used during compactions. We only count KV's that are older than all the scanners' read points.)

Returns:

the scan query matcher match code instance

Throws:

IOException - in case there is an internal consistency problem caused by a data corruption.

reset

public void reset()

Description copied from interface: ColumnTracker

Resets the Matcher

Specified by:

reset in interface ColumnTracker

doneWithColumn

public void doneWithColumn(Cell cell)

Description copied from interface: ColumnTracker

This method is used to inform the column tracker that we are done with this column. We may get this information from external filters or timestamp range and we then need to indicate this information to tracker. It is currently implemented for ExplicitColumnTracker.

Specified by:

doneWithColumn in interface ColumnTracker

getNextRowOrNextColumn

public ScanQueryMatcher.MatchCode getNextRowOrNextColumn(Cell cell)

Description copied from interface: ColumnTracker

Retrieve the MatchCode for the next row or column

Specified by:

getNextRowOrNextColumn in interface ColumnTracker

isDone

public boolean isDone(long timestamp)

Description copied from interface: ColumnTracker

Give the tracker a chance to declare it's done based on only the timestamp to allow an early out.

Specified by:

isDone in interface ColumnTracker

Returns:

true to early out based on timestamp.

beforeShipped

public void beforeShipped()
                   throws IOException

Description copied from interface: ShipperListener

The action that needs to be performed before Shipper.shipped() is performed

Specified by:

beforeShipped in interface ShipperListener

Throws:

IOException