Cells.java
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.cassandra.db.rows;
import java.nio.ByteBuffer;
import java.util.Comparator;
import java.util.Iterator;
import org.apache.cassandra.db.context.CounterContext;
import org.apache.cassandra.db.marshal.AbstractType;
import org.apache.cassandra.db.marshal.ValueAccessor;
import org.apache.cassandra.schema.ColumnMetadata;
import org.apache.cassandra.db.DeletionTime;
import org.apache.cassandra.db.partitions.PartitionStatisticsCollector;
/**
* Static methods to work on cells.
*/
public abstract class Cells
{
private Cells() {}
/**
* Collect statistics ont a given cell.
*
* @param cell the cell for which to collect stats.
* @param collector the stats collector.
*/
public static void collectStats(Cell<?> cell, PartitionStatisticsCollector collector)
{
collector.update(cell);
if (cell.isCounterCell())
collector.updateHasLegacyCounterShards(CounterCells.hasLegacyShards(cell));
}
/**
* Reconciles/merge two cells.
* <p>
* Note that this method assumes that the provided cells can meaningfully
* be reconciled together, that is that cell are for the same row and same
* column (and same cell path if the column is complex).
* <p>
* This method is commutative over it's cells arguments: {@code reconcile(a, b, n) == reconcile(b, a, n)}.
*
* @param c1 the first cell participating in the reconciliation.
* @param c2 the second cell participating in the reconciliation.
*
* @return a cell corresponding to the reconciliation of {@code c1} and {@code c2}.
* For non-counter cells, this will always be either {@code c1} or {@code c2}, but for
* counter cells this can be a newly allocated cell.
*/
public static Cell<?> reconcile(Cell<?> c1, Cell<?> c2)
{
if (c1 == null || c2 == null)
return c2 == null ? c1 : c2;
if (c1.isCounterCell() || c2.isCounterCell())
return resolveCounter(c1, c2);
return resolveRegular(c1, c2);
}
private static Cell<?> resolveRegular(Cell<?> left, Cell<?> right)
{
long leftTimestamp = left.timestamp();
long rightTimestamp = right.timestamp();
if (leftTimestamp != rightTimestamp)
return leftTimestamp > rightTimestamp ? left : right;
long leftLocalDeletionTime = left.localDeletionTime();
long rightLocalDeletionTime = right.localDeletionTime();
boolean leftIsExpiringOrTombstone = leftLocalDeletionTime != Cell.NO_DELETION_TIME;
boolean rightIsExpiringOrTombstone = rightLocalDeletionTime != Cell.NO_DELETION_TIME;
if (leftIsExpiringOrTombstone | rightIsExpiringOrTombstone)
{
// Tombstones always win reconciliation with live cells of the same timstamp
// CASSANDRA-14592: for consistency of reconciliation, regardless of system clock at time of reconciliation
// this requires us to treat expiring cells (which will become tombstones at some future date) the same wrt regular cells
if (leftIsExpiringOrTombstone != rightIsExpiringOrTombstone)
return leftIsExpiringOrTombstone ? left : right;
// for most historical consistency, we still prefer tombstones over expiring cells.
// While this leads to the an inconsistency over which is chosen
// (i.e. before expiry, the pure tombstone; after expiry, whichever is more recent)
// this inconsistency has no user-visible distinction, as at this point they are both logically tombstones
// (the only possible difference is the time at which the cells become purgeable)
boolean leftIsTombstone = !left.isExpiring(); // !isExpiring() == isTombstone(), but does not need to consider localDeletionTime()
boolean rightIsTombstone = !right.isExpiring();
if (leftIsTombstone != rightIsTombstone)
return leftIsTombstone ? left : right;
// ==> (leftIsExpiring && rightIsExpiring) or (leftIsTombstone && rightIsTombstone)
// if both are expiring, we do not want to consult the value bytes if we can avoid it, as like with C-14592
// the value bytes implicitly depend on the system time at reconciliation, as a
// would otherwise always win (unless it had an empty value), until it expired and was translated to a tombstone
if (leftLocalDeletionTime != rightLocalDeletionTime)
return leftLocalDeletionTime > rightLocalDeletionTime ? left : right;
}
return compareValues(left, right) >= 0 ? left : right;
}
private static Cell<?> resolveCounter(Cell<?> left, Cell<?> right)
{
long leftTimestamp = left.timestamp();
long rightTimestamp = right.timestamp();
boolean leftIsTombstone = left.isTombstone();
boolean rightIsTombstone = right.isTombstone();
if (leftIsTombstone | rightIsTombstone)
{
// No matter what the counter cell's timestamp is, a tombstone always takes precedence. See CASSANDRA-7346.
assert leftIsTombstone != rightIsTombstone;
return leftIsTombstone ? left : right;
}
ByteBuffer leftValue = left.buffer();
ByteBuffer rightValue = right.buffer();
// Handle empty values. Counters can't truly have empty values, but we can have a counter cell that temporarily
// has one on read if the column for the cell is not queried by the user due to the optimization of #10657. We
// thus need to handle this (see #11726 too).
boolean leftIsEmpty = !leftValue.hasRemaining();
boolean rightIsEmpty = !rightValue.hasRemaining();
if (leftIsEmpty || rightIsEmpty)
{
if (leftIsEmpty != rightIsEmpty)
return leftIsEmpty ? left : right;
return leftTimestamp > rightTimestamp ? left : right;
}
ByteBuffer merged = CounterContext.instance().merge(leftValue, rightValue);
long timestamp = Math.max(leftTimestamp, rightTimestamp);
// We save allocating a new cell object if it turns out that one cell was
// a complete superset of the other
if (merged == leftValue && timestamp == leftTimestamp)
return left;
else if (merged == rightValue && timestamp == rightTimestamp)
return right;
else // merge clocks and timestamps.
return new BufferCell(left.column(), timestamp, Cell.NO_TTL, Cell.NO_DELETION_TIME, merged, left.path());
}
/**
* Adds to the builder a representation of the given existing cell that, when merged/reconciled with the given
* update cell, produces the same result as merging the original with the update.
* <p>
* For simple cells that is either the original cell (if still live), or nothing (if shadowed).
*
* @param existing the pre-existing cell, the one that is updated.
* @param update the newly added cell, the update. This can be {@code null} out
* of convenience, in which case this function simply copy {@code existing} to
* {@code writer}.
* @param deletion the deletion time that applies to the cells being considered.
* This deletion time may delete both {@code existing} or {@code update}.
* @param builder the row builder to which the result of the filtering is written.
*/
public static void addNonShadowed(Cell<?> existing,
Cell<?> update,
DeletionTime deletion,
Row.Builder builder)
{
if (deletion.deletes(existing))
return;
Cell<?> reconciled = reconcile(existing, update);
if (reconciled != update)
builder.addCell(existing);
}
/**
* Adds to the builder a representation of the given existing cell that, when merged/reconciled with the given
* update cell, produces the same result as merging the original with the update.
* <p>
* For simple cells that is either the original cell (if still live), or nothing (if shadowed).
*
* @param column the complex column the cells are for.
* @param existing the pre-existing cells, the ones that are updated.
* @param update the newly added cells, the update. This can be {@code null} out
* of convenience, in which case this function simply copy the cells from
* {@code existing} to {@code writer}.
* @param deletion the deletion time that applies to the cells being considered.
* This deletion time may delete both {@code existing} or {@code update}.
* @param builder the row builder to which the result of the filtering is written.
*/
public static void addNonShadowedComplex(ColumnMetadata column,
Iterator<Cell<?>> existing,
Iterator<Cell<?>> update,
DeletionTime deletion,
Row.Builder builder)
{
Comparator<CellPath> comparator = column.cellPathComparator();
Cell<?> nextExisting = getNext(existing);
Cell<?> nextUpdate = getNext(update);
while (nextExisting != null)
{
int cmp = nextUpdate == null ? -1 : comparator.compare(nextExisting.path(), nextUpdate.path());
if (cmp < 0)
{
addNonShadowed(nextExisting, null, deletion, builder);
nextExisting = getNext(existing);
}
else if (cmp == 0)
{
addNonShadowed(nextExisting, nextUpdate, deletion, builder);
nextExisting = getNext(existing);
nextUpdate = getNext(update);
}
else
{
nextUpdate = getNext(update);
}
}
}
private static Cell<?> getNext(Iterator<Cell<?>> iterator)
{
return iterator == null || !iterator.hasNext() ? null : iterator.next();
}
private static <L, R> int compareValues(Cell<L> left, Cell<R> right)
{
return ValueAccessor.compare(left.value(), left.accessor(), right.value(), right.accessor());
}
public static <L, R> boolean valueEqual(Cell<L> left, Cell<R> right)
{
return ValueAccessor.equals(left.value(), left.accessor(), right.value(), right.accessor());
}
public static <T, V> T composeValue(Cell<V> cell, AbstractType<T> type)
{
return type.compose(cell.value(), cell.accessor());
}
public static <V> String valueString(Cell<V> cell, AbstractType<?> type)
{
return type.getString(cell.value(), cell.accessor());
}
public static <V> String valueString(Cell<V> cell)
{
return valueString(cell, cell.column().type);
}
}