org.apache.cassandra.db

Class RangeTombstone.Tracker

java.lang.Object

org.apache.cassandra.db.RangeTombstone.Tracker

Enclosing class:

RangeTombstone

public static class RangeTombstone.Tracker
extends java.lang.Object

Tracks opened RangeTombstones when iterating over a partition.

This tracker must be provided all the atoms of a given partition in order (to the update method). Given this, it keeps enough information to be able to decide if one of an atom is deleted (shadowed) by a previously open RT. One the tracker can prove a given range tombstone cannot be useful anymore (that is, as soon as we've seen an atom that is after the end of that RT), it discards this RT. In other words, the maximum memory used by this object should be proportional to the maximum number of RT that can be simultaneously open (and this should fairly low in practice).

Constructor Summary

Constructors

Constructor and Description
Tracker(java.util.Comparator<Composite> comparator) Creates a new tracker given the table comparator.

Method Summary

Modifier and Type	Method and Description
boolean	hasUnwrittenTombstones()
boolean	isDeleted(Cell cell) Tests if the provided column is deleted by one of the tombstone tracked by this tracker.
boolean	update(OnDiskAtom atom, boolean isExpired) Update this tracker given an atom.
long	writeOpenedMarkers(Composite startPos, DataOutputPlus out, OnDiskAtom.SerializerForWriting atomSerializer) Computes the RangeTombstone that are needed at the beginning of an index block starting with firstColumn.
long	writeUnwrittenTombstones(DataOutputPlus out, OnDiskAtom.SerializerForWriting atomSerializer) Writes out all tombstones that have been accepted after the previous call of this method.
int	writtenAtom() The total number of atoms written by calls to the above methods.

Methods inherited from class java.lang.Object

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

Constructor Detail

Tracker

public Tracker(java.util.Comparator<Composite> comparator)

Creates a new tracker given the table comparator.

Parameters:

comparator - the comparator for the table this will track atoms for. The tracker assumes that atoms will be later provided to the tracker in comparator order.

Method Detail

writeOpenedMarkers

public long writeOpenedMarkers(Composite startPos,
                               DataOutputPlus out,
                               OnDiskAtom.SerializerForWriting atomSerializer)
                        throws java.io.IOException

Computes the RangeTombstone that are needed at the beginning of an index block starting with firstColumn.

Returns:

the total serialized size of said tombstones and write them to out it if isn't null.

Throws:

java.io.IOException

writeUnwrittenTombstones

public long writeUnwrittenTombstones(DataOutputPlus out,
                                     OnDiskAtom.SerializerForWriting atomSerializer)
                              throws java.io.IOException

Writes out all tombstones that have been accepted after the previous call of this method. Tombstones are not written immediately to permit redundant ones to be skipped.

Returns:

the serialized size of written tombstones

Throws:

java.io.IOException

writtenAtom

public int writtenAtom()

The total number of atoms written by calls to the above methods.

update

public boolean update(OnDiskAtom atom,
                      boolean isExpired)

Update this tracker given an atom.

This method first test if some range tombstone can be discarded due to the knowledge of that new atom. Then, if it's a range tombstone, it adds it to the tracker.

Note that this method should be called on *every* atom of a partition for the tracker to work as efficiently as possible (#9486).

isDeleted

public boolean isDeleted(Cell cell)

Tests if the provided column is deleted by one of the tombstone tracked by this tracker.

This method should be called on columns in the same order than for the update() method. Note that this method does not update the tracker so the update() method should still be called on column (it doesn't matter if update is called before or after this call).

hasUnwrittenTombstones

public boolean hasUnwrittenTombstones()