/*
* 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.partitions;
import java.io.IOException;
import java.nio.ByteBuffer;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;
import com.google.common.collect.Iterables;
import com.google.common.collect.Lists;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.apache.cassandra.config.CFMetaData;
import org.apache.cassandra.config.ColumnDefinition;
import org.apache.cassandra.db.*;
import org.apache.cassandra.db.filter.ColumnFilter;
import org.apache.cassandra.db.rows.*;
import org.apache.cassandra.io.util.*;
import org.apache.cassandra.net.MessagingService;
import org.apache.cassandra.utils.FBUtilities;
import org.apache.cassandra.utils.btree.BTree;
import org.apache.cassandra.utils.btree.UpdateFunction;
/**
* Stores updates made on a partition.
* <p>
* A PartitionUpdate object requires that all writes/additions are performed before we
* try to read the updates (attempts to write to the PartitionUpdate after a read method
* has been called will result in an exception being thrown). In other words, a Partition
* is mutable while it's written but becomes immutable as soon as it is read.
* <p>
* A typical usage is to create a new update ({@code new PartitionUpdate(metadata, key, columns, capacity)})
* and then add rows and range tombstones through the {@code add()} methods (the partition
* level deletion time can also be set with {@code addPartitionDeletion()}). However, there
* is also a few static helper constructor methods for special cases ({@code emptyUpdate()},
* {@code fullPartitionDelete} and {@code singleRowUpdate}).
*/
public class PartitionUpdate extends AbstractBTreePartition
{
protected static final Logger logger = LoggerFactory.getLogger(PartitionUpdate.class);
public static final PartitionUpdateSerializer serializer = new PartitionUpdateSerializer();
private final int createdAtInSec = FBUtilities.nowInSeconds();
// Records whether this update is "built", i.e. if the build() method has been called, which
// happens when the update is read. Further writing is then rejected though a manual call
// to allowNewUpdates() allow new writes. We could make that more implicit but only triggers
// really requires that so we keep it simple for now).
private boolean isBuilt;
private boolean canReOpen = true;
private Holder holder;
private BTree.Builder<Row> rowBuilder;
private MutableDeletionInfo deletionInfo;
private final boolean canHaveShadowedData;
private PartitionUpdate(CFMetaData metadata,
DecoratedKey key,
PartitionColumns columns,
MutableDeletionInfo deletionInfo,
int initialRowCapacity,
boolean canHaveShadowedData)
{
super(metadata, key);
this.deletionInfo = deletionInfo;
this.holder = new Holder(columns, BTree.empty(), deletionInfo, Rows.EMPTY_STATIC_ROW, EncodingStats.NO_STATS);
this.canHaveShadowedData = canHaveShadowedData;
rowBuilder = builder(initialRowCapacity);
}
private PartitionUpdate(CFMetaData metadata,
DecoratedKey key,
Holder holder,
MutableDeletionInfo deletionInfo,
boolean canHaveShadowedData)
{
super(metadata, key);
this.holder = holder;
this.deletionInfo = deletionInfo;
this.isBuilt = true;
this.canHaveShadowedData = canHaveShadowedData;
}
public PartitionUpdate(CFMetaData metadata,
DecoratedKey key,
PartitionColumns columns,
int initialRowCapacity)
{
this(metadata, key, columns, MutableDeletionInfo.live(), initialRowCapacity, true);
}
public PartitionUpdate(CFMetaData metadata,
ByteBuffer key,
PartitionColumns columns,
int initialRowCapacity)
{
this(metadata,
metadata.decorateKey(key),
columns,
initialRowCapacity);
}
/**
* Creates a empty immutable partition update.
*
* @param metadata the metadata for the created update.
* @param key the partition key for the created update.
*
* @return the newly created empty (and immutable) update.
*/
public static PartitionUpdate emptyUpdate(CFMetaData metadata, DecoratedKey key)
{
MutableDeletionInfo deletionInfo = MutableDeletionInfo.live();
Holder holder = new Holder(PartitionColumns.NONE, BTree.empty(), deletionInfo, Rows.EMPTY_STATIC_ROW, EncodingStats.NO_STATS);
return new PartitionUpdate(metadata, key, holder, deletionInfo, false);
}
/**
* Creates an immutable partition update that entirely deletes a given partition.
*
* @param metadata the metadata for the created update.
* @param key the partition key for the partition that the created update should delete.
* @param timestamp the timestamp for the deletion.
* @param nowInSec the current time in seconds to use as local deletion time for the partition deletion.
*
* @return the newly created partition deletion update.
*/
public static PartitionUpdate fullPartitionDelete(CFMetaData metadata, DecoratedKey key, long timestamp, int nowInSec)
{
MutableDeletionInfo deletionInfo = new MutableDeletionInfo(timestamp, nowInSec);
Holder holder = new Holder(PartitionColumns.NONE, BTree.empty(), deletionInfo, Rows.EMPTY_STATIC_ROW, EncodingStats.NO_STATS);
return new PartitionUpdate(metadata, key, holder, deletionInfo, false);
}
/**
* Creates an immutable partition update that contains a single row update.
*
* @param metadata the metadata for the created update.
* @param key the partition key for the partition to update.
* @param row the row for the update.
*
* @return the newly created partition update containing only {@code row}.
*/
public static PartitionUpdate singleRowUpdate(CFMetaData metadata, DecoratedKey key, Row row)
{
MutableDeletionInfo deletionInfo = MutableDeletionInfo.live();
if (row.isStatic())
{
Holder holder = new Holder(new PartitionColumns(Columns.from(row.columns()), Columns.NONE), BTree.empty(), deletionInfo, row, EncodingStats.NO_STATS);
return new PartitionUpdate(metadata, key, holder, deletionInfo, false);
}
else
{
Holder holder = new Holder(new PartitionColumns(Columns.NONE, Columns.from(row.columns())), BTree.singleton(row), deletionInfo, Rows.EMPTY_STATIC_ROW, EncodingStats.NO_STATS);
return new PartitionUpdate(metadata, key, holder, deletionInfo, false);
}
}
/**
* Creates an immutable partition update that contains a single row update.
*
* @param metadata the metadata for the created update.
* @param key the partition key for the partition to update.
* @param row the row for the update.
*
* @return the newly created partition update containing only {@code row}.
*/
public static PartitionUpdate singleRowUpdate(CFMetaData metadata, ByteBuffer key, Row row)
{
return singleRowUpdate(metadata, metadata.decorateKey(key), row);
}
/**
* Turns the given iterator into an update.
*
* Warning: this method does not close the provided iterator, it is up to
* the caller to close it.
*/
public static PartitionUpdate fromIterator(UnfilteredRowIterator iterator)
{
Holder holder = build(iterator, 16);
MutableDeletionInfo deletionInfo = (MutableDeletionInfo) holder.deletionInfo;
return new PartitionUpdate(iterator.metadata(), iterator.partitionKey(), holder, deletionInfo, false);
}
public static PartitionUpdate fromIterator(RowIterator iterator)
{
MutableDeletionInfo deletionInfo = MutableDeletionInfo.live();
Holder holder = build(iterator, deletionInfo, true, 16);
return new PartitionUpdate(iterator.metadata(), iterator.partitionKey(), holder, deletionInfo, false);
}
protected boolean canHaveShadowedData()
{
return canHaveShadowedData;
}
/**
* Deserialize a partition update from a provided byte buffer.
*
* @param bytes the byte buffer that contains the serialized update.
* @param version the version with which the update is serialized.
* @param key the partition key for the update. This is only used if {@code version < 3.0}
* and can be {@code null} otherwise.
*
* @return the deserialized update or {@code null} if {@code bytes == null}.
*/
public static PartitionUpdate fromBytes(ByteBuffer bytes, int version, DecoratedKey key)
{
if (bytes == null)
return null;
try
{
return serializer.deserialize(new DataInputBuffer(bytes, true),
version,
SerializationHelper.Flag.LOCAL,
version < MessagingService.VERSION_30 ? key : null);
}
catch (IOException e)
{
throw new RuntimeException(e);
}
}
/**
* Serialize a partition update as a byte buffer.
*
* @param update the partition update to serialize.
* @param version the version to serialize the update into.
*
* @return a newly allocated byte buffer containing the serialized update.
*/
public static ByteBuffer toBytes(PartitionUpdate update, int version)
{
try (DataOutputBuffer out = new DataOutputBuffer())
{
serializer.serialize(update, out, version);
return ByteBuffer.wrap(out.getData(), 0, out.getLength());
}
catch (IOException e)
{
throw new RuntimeException(e);
}
}
/**
* Creates a partition update that entirely deletes a given partition.
*
* @param metadata the metadata for the created update.
* @param key the partition key for the partition that the created update should delete.
* @param timestamp the timestamp for the deletion.
* @param nowInSec the current time in seconds to use as local deletion time for the partition deletion.
*
* @return the newly created partition deletion update.
*/
public static PartitionUpdate fullPartitionDelete(CFMetaData metadata, ByteBuffer key, long timestamp, int nowInSec)
{
return fullPartitionDelete(metadata, metadata.decorateKey(key), timestamp, nowInSec);
}
/**
* Merges the provided updates, yielding a new update that incorporates all those updates.
*
* @param updates the collection of updates to merge. This shouldn't be empty.
*
* @return a partition update that include (merge) all the updates from {@code updates}.
*/
public static PartitionUpdate merge(List<PartitionUpdate> updates)
{
assert !updates.isEmpty();
final int size = updates.size();
if (size == 1)
return Iterables.getOnlyElement(updates);
int nowInSecs = FBUtilities.nowInSeconds();
List<UnfilteredRowIterator> asIterators = Lists.transform(updates, AbstractBTreePartition::unfilteredIterator);
return fromIterator(UnfilteredRowIterators.merge(asIterators, nowInSecs));
}
/**
* Modify this update to set every timestamp for live data to {@code newTimestamp} and
* every deletion timestamp to {@code newTimestamp - 1}.
*
* There is no reason to use that expect on the Paxos code path, where we need ensure that
* anything inserted use the ballot timestamp (to respect the order of update decided by
* the Paxos algorithm). We use {@code newTimestamp - 1} for deletions because tombstones
* always win on timestamp equality and we don't want to delete our own insertions
* (typically, when we overwrite a collection, we first set a complex deletion to delete the
* previous collection before adding new elements. If we were to set that complex deletion
* to the same timestamp that the new elements, it would delete those elements). And since
* tombstones always wins on timestamp equality, using -1 guarantees our deletion will still
* delete anything from a previous update.
*/
public void updateAllTimestamp(long newTimestamp)
{
Holder holder = holder();
deletionInfo.updateAllTimestamp(newTimestamp - 1);
Object[] tree = BTree.<Row>transformAndFilter(holder.tree, (x) -> x.updateAllTimestamp(newTimestamp));
Row staticRow = holder.staticRow.updateAllTimestamp(newTimestamp);
EncodingStats newStats = EncodingStats.Collector.collect(staticRow, BTree.<Row>iterator(tree), deletionInfo);
this.holder = new Holder(holder.columns, tree, deletionInfo, staticRow, newStats);
}
/**
* The number of "operations" contained in the update.
* <p>
* This is used by {@code Memtable} to approximate how much work this update does. In practice, this
* count how many rows are updated and how many ranges are deleted by the partition update.
*
* @return the number of "operations" performed by the update.
*/
public int operationCount()
{
return rowCount()
+ (staticRow().isEmpty() ? 0 : 1)
+ deletionInfo.rangeCount()
+ (deletionInfo.getPartitionDeletion().isLive() ? 0 : 1);
}
/**
* The size of the data contained in this update.
*
* @return the size of the data contained in this update.
*/
public int dataSize()
{
int size = 0;
for (Row row : this)
{
size += row.clustering().dataSize();
for (ColumnData cd : row)
size += cd.dataSize();
}
return size;
}
@Override
public PartitionColumns columns()
{
// The superclass implementation calls holder(), but that triggers a build of the PartitionUpdate. But since
// the columns are passed to the ctor, we know the holder always has the proper columns even if it doesn't have
// the built rows yet, so just bypass the holder() method.
return holder.columns;
}
protected Holder holder()
{
maybeBuild();
return holder;
}
public EncodingStats stats()
{
return holder().stats;
}
/**
* If a partition update has been read (and is thus unmodifiable), a call to this method
* makes the update modifiable again.
* <p>
* Please note that calling this method won't result in optimal behavior in the sense that
* even if very little is added to the update after this call, the whole update will be sorted
* again on read. This should thus be used sparingly (and if it turns that we end up using
* this often, we should consider optimizing the behavior).
*/
public synchronized void allowNewUpdates()
{
if (!canReOpen)
throw new IllegalStateException("You cannot do more updates on collectCounterMarks has been called");
// This is synchronized to make extra sure things work properly even if this is
// called concurrently with sort() (which should be avoided in the first place, but
// better safe than sorry).
isBuilt = false;
if (rowBuilder == null)
rowBuilder = builder(16);
}
private BTree.Builder<Row> builder(int initialCapacity)
{
return BTree.<Row>builder(metadata.comparator, initialCapacity)
.setQuickResolver((a, b) ->
Rows.merge(a, b, createdAtInSec));
}
/**
* Returns an iterator that iterates over the rows of this update in clustering order.
* <p>
* Note that this might trigger a sorting of the update, and as such the update will not
* be modifiable anymore after this call.
*
* @return an iterator over the rows of this update.
*/
@Override
public Iterator<Row> iterator()
{
maybeBuild();
return super.iterator();
}
@Override
public SliceableUnfilteredRowIterator sliceableUnfilteredIterator(ColumnFilter columns, boolean reversed)
{
maybeBuild();
return super.sliceableUnfilteredIterator(columns, reversed);
}
/**
* Validates the data contained in this update.
*
* @throws org.apache.cassandra.serializers.MarshalException if some of the data contained in this update is corrupted.
*/
public void validate()
{
for (Row row : this)
{
metadata().comparator.validate(row.clustering());
for (ColumnData cd : row)
cd.validate();
}
}
/**
* The maximum timestamp used in this update.
*
* @return the maximum timestamp used in this update.
*/
public long maxTimestamp()
{
maybeBuild();
long maxTimestamp = deletionInfo.maxTimestamp();
for (Row row : this)
{
maxTimestamp = Math.max(maxTimestamp, row.primaryKeyLivenessInfo().timestamp());
for (ColumnData cd : row)
{
if (cd.column().isSimple())
{
maxTimestamp = Math.max(maxTimestamp, ((Cell)cd).timestamp());
}
else
{
ComplexColumnData complexData = (ComplexColumnData)cd;
maxTimestamp = Math.max(maxTimestamp, complexData.complexDeletion().markedForDeleteAt());
for (Cell cell : complexData)
maxTimestamp = Math.max(maxTimestamp, cell.timestamp());
}
}
}
return maxTimestamp;
}
/**
* For an update on a counter table, returns a list containing a {@code CounterMark} for
* every counter contained in the update.
*
* @return a list with counter marks for every counter in this update.
*/
public List<CounterMark> collectCounterMarks()
{
assert metadata().isCounter();
maybeBuild();
// We will take aliases on the rows of this update, and update them in-place. So we should be sure the
// update is now immutable for all intent and purposes.
canReOpen = false;
List<CounterMark> marks = new ArrayList<>();
addMarksForRow(staticRow(), marks);
for (Row row : this)
addMarksForRow(row, marks);
return marks;
}
private void addMarksForRow(Row row, List<CounterMark> marks)
{
for (Cell cell : row.cells())
{
if (cell.isCounterCell())
marks.add(new CounterMark(row, cell.column(), cell.path()));
}
}
private void assertNotBuilt()
{
if (isBuilt)
throw new IllegalStateException("An update should not be written again once it has been read");
}
public void addPartitionDeletion(DeletionTime deletionTime)
{
assertNotBuilt();
deletionInfo.add(deletionTime);
}
public void add(RangeTombstone range)
{
assertNotBuilt();
deletionInfo.add(range, metadata.comparator);
}
/**
* Adds a row to this update.
*
* There is no particular assumption made on the order of row added to a partition update. It is further
* allowed to add the same row (more precisely, multiple row objects for the same clustering).
*
* Note however that the columns contained in the added row must be a subset of the columns used when
* creating this update.
*
* @param row the row to add.
*/
public void add(Row row)
{
if (row.isEmpty())
return;
assertNotBuilt();
if (row.isStatic())
{
// this assert is expensive, and possibly of limited value; we should consider removing it
// or introducing a new class of assertions for test purposes
assert columns().statics.containsAll(row.columns()) : columns().statics + " is not superset of " + row.columns();
Row staticRow = holder.staticRow.isEmpty()
? row
: Rows.merge(holder.staticRow, row, createdAtInSec);
holder = new Holder(holder.columns, holder.tree, holder.deletionInfo, staticRow, holder.stats);
}
else
{
// this assert is expensive, and possibly of limited value; we should consider removing it
// or introducing a new class of assertions for test purposes
assert columns().regulars.containsAll(row.columns()) : columns().regulars + " is not superset of " + row.columns();
rowBuilder.add(row);
}
}
private void maybeBuild()
{
if (isBuilt)
return;
build();
}
private synchronized void build()
{
if (isBuilt)
return;
Holder holder = this.holder;
Object[] cur = holder.tree;
Object[] add = rowBuilder.build();
Object[] merged = BTree.<Row>merge(cur, add, metadata.comparator,
UpdateFunction.Simple.of((a, b) -> Rows.merge(a, b, createdAtInSec)));
assert deletionInfo == holder.deletionInfo;
EncodingStats newStats = EncodingStats.Collector.collect(holder.staticRow, BTree.<Row>iterator(merged), deletionInfo);
this.holder = new Holder(holder.columns, merged, holder.deletionInfo, holder.staticRow, newStats);
rowBuilder = null;
isBuilt = true;
}
@Override
public String toString()
{
if (isBuilt)
return super.toString();
// We intentionally override AbstractBTreePartition#toString() to avoid iterating over the rows in the
// partition, which can result in build() being triggered and lead to errors if the PartitionUpdate is later
// modified.
StringBuilder sb = new StringBuilder();
sb.append(String.format("[%s.%s] key=%s columns=%s",
metadata.ksName,
metadata.cfName,
metadata.getKeyValidator().getString(partitionKey().getKey()),
columns()));
sb.append("\n deletionInfo=").append(deletionInfo);
sb.append(" (not built)");
return sb.toString();
}
public static class PartitionUpdateSerializer
{
public void serialize(PartitionUpdate update, DataOutputPlus out, int version) throws IOException
{
try (UnfilteredRowIterator iter = update.sliceableUnfilteredIterator())
{
assert !iter.isReverseOrder();
if (version < MessagingService.VERSION_30)
{
LegacyLayout.serializeAsLegacyPartition(null, iter, out, version);
}
else
{
CFMetaData.serializer.serialize(update.metadata(), out, version);
UnfilteredRowIteratorSerializer.serializer.serialize(iter, null, out, version, update.rowCount());
}
}
}
public PartitionUpdate deserialize(DataInputPlus in, int version, SerializationHelper.Flag flag, ByteBuffer key) throws IOException
{
if (version >= MessagingService.VERSION_30)
{
assert key == null; // key is only there for the old format
return deserialize30(in, version, flag);
}
else
{
assert key != null;
return deserializePre30(in, version, flag, key);
}
}
// Used to share same decorated key between updates.
public PartitionUpdate deserialize(DataInputPlus in, int version, SerializationHelper.Flag flag, DecoratedKey key) throws IOException
{
if (version >= MessagingService.VERSION_30)
{
return deserialize30(in, version, flag);
}
else
{
assert key != null;
return deserializePre30(in, version, flag, key.getKey());
}
}
private static PartitionUpdate deserialize30(DataInputPlus in, int version, SerializationHelper.Flag flag) throws IOException
{
CFMetaData metadata = CFMetaData.serializer.deserialize(in, version);
UnfilteredRowIteratorSerializer.Header header = UnfilteredRowIteratorSerializer.serializer.deserializeHeader(metadata, null, in, version, flag);
if (header.isEmpty)
return emptyUpdate(metadata, header.key);
assert !header.isReversed;
assert header.rowEstimate >= 0;
MutableDeletionInfo.Builder deletionBuilder = MutableDeletionInfo.builder(header.partitionDeletion, metadata.comparator, false);
BTree.Builder<Row> rows = BTree.builder(metadata.comparator, header.rowEstimate);
rows.auto(false);
try (UnfilteredRowIterator partition = UnfilteredRowIteratorSerializer.serializer.deserialize(in, version, metadata, flag, header))
{
while (partition.hasNext())
{
Unfiltered unfiltered = partition.next();
if (unfiltered.kind() == Unfiltered.Kind.ROW)
rows.add((Row)unfiltered);
else
deletionBuilder.add((RangeTombstoneMarker)unfiltered);
}
}
MutableDeletionInfo deletionInfo = deletionBuilder.build();
return new PartitionUpdate(metadata,
header.key,
new Holder(header.sHeader.columns(), rows.build(), deletionInfo, header.staticRow, header.sHeader.stats()),
deletionInfo,
false);
}
private static PartitionUpdate deserializePre30(DataInputPlus in, int version, SerializationHelper.Flag flag, ByteBuffer key) throws IOException
{
try (UnfilteredRowIterator iterator = LegacyLayout.deserializeLegacyPartition(in, version, flag, key))
{
assert iterator != null; // This is only used in mutation, and mutation have never allowed "null" column families
return PartitionUpdate.fromIterator(iterator);
}
}
public long serializedSize(PartitionUpdate update, int version)
{
try (UnfilteredRowIterator iter = update.sliceableUnfilteredIterator())
{
if (version < MessagingService.VERSION_30)
return LegacyLayout.serializedSizeAsLegacyPartition(null, iter, version);
return CFMetaData.serializer.serializedSize(update.metadata(), version)
+ UnfilteredRowIteratorSerializer.serializer.serializedSize(iter, null, version, update.rowCount());
}
}
}
/**
* A counter mark is basically a pointer to a counter update inside this partition update. That pointer allows
* us to update the counter value based on the pre-existing value read during the read-before-write that counters
* do. See {@link CounterMutation} to understand how this is used.
*/
public static class CounterMark
{
private final Row row;
private final ColumnDefinition column;
private final CellPath path;
private CounterMark(Row row, ColumnDefinition column, CellPath path)
{
this.row = row;
this.column = column;
this.path = path;
}
public Clustering clustering()
{
return row.clustering();
}
public ColumnDefinition column()
{
return column;
}
public CellPath path()
{
return path;
}
public ByteBuffer value()
{
return path == null
? row.getCell(column).value()
: row.getCell(column, path).value();
}
public void setValue(ByteBuffer value)
{
// This is a bit of a giant hack as this is the only place where we mutate a Row object. This makes it more efficient
// for counters however and this won't be needed post-#6506 so that's probably fine.
assert row instanceof BTreeRow;
((BTreeRow)row).setValue(column, path, value);
}
}
}