org.apache.cassandra.service.pager.PagingState Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of cassandra-all Show documentation
Show all versions of cassandra-all Show documentation
The Apache Cassandra Project develops a highly scalable second-generation distributed database, bringing together Dynamo's fully distributed design and Bigtable's ColumnFamily-based data model.
/*
* 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.service.pager;
import java.io.IOException;
import java.nio.ByteBuffer;
import java.util.*;
import com.google.common.annotations.VisibleForTesting;
import com.google.common.primitives.Ints;
import org.apache.cassandra.db.Clustering;
import org.apache.cassandra.db.marshal.AbstractType;
import org.apache.cassandra.db.marshal.ByteArrayAccessor;
import org.apache.cassandra.db.marshal.ByteBufferAccessor;
import org.apache.cassandra.db.marshal.BytesType;
import org.apache.cassandra.db.marshal.CompositeType;
import org.apache.cassandra.db.rows.Cell;
import org.apache.cassandra.db.rows.Row;
import org.apache.cassandra.io.util.DataInputBuffer;
import org.apache.cassandra.io.util.DataOutputBuffer;
import org.apache.cassandra.io.util.DataOutputBufferFixed;
import org.apache.cassandra.net.MessagingService;
import org.apache.cassandra.schema.TableMetadata;
import org.apache.cassandra.transport.ProtocolException;
import org.apache.cassandra.transport.ProtocolVersion;
import static org.apache.cassandra.db.TypeSizes.sizeof;
import static org.apache.cassandra.db.TypeSizes.sizeofUnsignedVInt;
import static org.apache.cassandra.utils.ByteBufferUtil.*;
import static org.apache.cassandra.utils.vint.VIntCoding.computeUnsignedVIntSize;
import static org.apache.cassandra.utils.vint.VIntCoding.getUnsignedVInt;
@SuppressWarnings("WeakerAccess")
public class PagingState
{
public final ByteBuffer partitionKey; // Can be null for single partition queries.
public final RowMark rowMark; // Can be null if not needed.
public final int remaining;
public final int remainingInPartition;
public PagingState(ByteBuffer partitionKey, RowMark rowMark, int remaining, int remainingInPartition)
{
this.partitionKey = partitionKey;
this.rowMark = rowMark;
this.remaining = remaining;
this.remainingInPartition = remainingInPartition;
}
public ByteBuffer serialize(ProtocolVersion protocolVersion)
{
assert rowMark == null || protocolVersion == rowMark.protocolVersion;
try
{
return protocolVersion.isGreaterThan(ProtocolVersion.V3) ? modernSerialize() : legacySerialize(true);
}
catch (IOException e)
{
throw new RuntimeException(e);
}
}
public int serializedSize(ProtocolVersion protocolVersion)
{
assert rowMark == null || protocolVersion == rowMark.protocolVersion;
return protocolVersion.isGreaterThan(ProtocolVersion.V3) ? modernSerializedSize() : legacySerializedSize(true);
}
/**
* It's possible to receive a V3 paging state on a V4 client session, and vice versa - so we cannot
* blindly rely on the protocol version provided. We must verify first that the buffer indeed contains
* a paging state that adheres to the protocol version provided, or, if not - see if it is in a different
* version, in which case we try the other format.
*/
public static PagingState deserialize(ByteBuffer bytes, ProtocolVersion protocolVersion)
{
if (bytes == null)
return null;
try
{
/*
* We can't just attempt to deser twice, as we risk to misinterpet short/vint
* lengths and allocate huge byte arrays for readWithVIntLength() or,
* to a lesser extent, readWithShortLength()
*/
if (protocolVersion.isGreaterThan(ProtocolVersion.V3))
{
if (isModernSerialized(bytes)) return modernDeserialize(bytes, protocolVersion);
if (isLegacySerialized(bytes)) return legacyDeserialize(bytes, ProtocolVersion.V3);
}
if (protocolVersion.isSmallerThan(ProtocolVersion.V4))
{
if (isLegacySerialized(bytes)) return legacyDeserialize(bytes, protocolVersion);
if (isModernSerialized(bytes)) return modernDeserialize(bytes, ProtocolVersion.V4);
}
}
catch (IOException e)
{
throw new ProtocolException("Invalid value for the paging state");
}
throw new ProtocolException("Invalid value for the paging state");
}
/*
* Modern serde (> VERSION_3)
*/
@SuppressWarnings({ "resource", "RedundantSuppression" })
private ByteBuffer modernSerialize() throws IOException
{
DataOutputBuffer out = new DataOutputBufferFixed(modernSerializedSize());
writeWithVIntLength(null == partitionKey ? EMPTY_BYTE_BUFFER : partitionKey, out);
writeWithVIntLength(null == rowMark ? EMPTY_BYTE_BUFFER : rowMark.mark, out);
out.writeUnsignedVInt(remaining);
out.writeUnsignedVInt(remainingInPartition);
return out.buffer(false);
}
@VisibleForTesting
static boolean isModernSerialized(ByteBuffer bytes)
{
int index = bytes.position();
int limit = bytes.limit();
int partitionKeyLen = toIntExact(getUnsignedVInt(bytes, index, limit));
if (partitionKeyLen < 0)
return false;
index = addNonNegative(index, computeUnsignedVIntSize(partitionKeyLen), partitionKeyLen);
if (index >= limit || index < 0)
return false;
int rowMarkerLen = toIntExact(getUnsignedVInt(bytes, index, limit));
if (rowMarkerLen < 0)
return false;
index = addNonNegative(index, computeUnsignedVIntSize(rowMarkerLen), rowMarkerLen);
if (index >= limit || index < 0)
return false;
int remaining = toIntExact(getUnsignedVInt(bytes, index, limit));
if (remaining < 0)
return false;
index = addNonNegative(index, computeUnsignedVIntSize(remaining));
if (index >= limit || index < 0)
return false;
long remainingInPartition = getUnsignedVInt(bytes, index, limit);
if (remainingInPartition < 0)
return false;
index = addNonNegative(index, computeUnsignedVIntSize(remainingInPartition));
return index == limit;
}
// Following operations are similar to Math.{addExact/toIntExact}, but without using exceptions for control flow.
// Since we're operating non-negative numbers, we can use -1 return value as an error code.
private static int addNonNegative(int x, int y)
{
int sum = x + y;
if (sum < 0)
return -1;
return sum;
}
private static int addNonNegative(int x, int y, int z)
{
int sum = x + y;
if (sum < 0)
return -1;
sum += z;
if (sum < 0)
return -1;
return sum;
}
private static int toIntExact(long value)
{
if ((int)value != value)
return -1;
return (int)value;
}
@SuppressWarnings({ "resource", "RedundantSuppression" })
private static PagingState modernDeserialize(ByteBuffer bytes, ProtocolVersion protocolVersion) throws IOException
{
if (protocolVersion.isSmallerThan(ProtocolVersion.V4))
throw new IllegalArgumentException();
DataInputBuffer in = new DataInputBuffer(bytes, false);
ByteBuffer partitionKey = readWithVIntLength(in);
ByteBuffer rawMark = readWithVIntLength(in);
int remaining = Ints.checkedCast(in.readUnsignedVInt());
int remainingInPartition = Ints.checkedCast(in.readUnsignedVInt());
return new PagingState(partitionKey.hasRemaining() ? partitionKey : null,
rawMark.hasRemaining() ? new RowMark(rawMark, protocolVersion) : null,
remaining,
remainingInPartition);
}
private int modernSerializedSize()
{
return serializedSizeWithVIntLength(null == partitionKey ? EMPTY_BYTE_BUFFER : partitionKey)
+ serializedSizeWithVIntLength(null == rowMark ? EMPTY_BYTE_BUFFER : rowMark.mark)
+ sizeofUnsignedVInt(remaining)
+ sizeofUnsignedVInt(remainingInPartition);
}
/*
* Legacy serde (< VERSION_4)
*
* There are two versions of legacy PagingState format - one used by 2.1/2.2 and one used by 3.0+.
* The latter includes remainingInPartition count, while the former doesn't.
*/
@VisibleForTesting
@SuppressWarnings({ "resource", "RedundantSuppression" })
ByteBuffer legacySerialize(boolean withRemainingInPartition) throws IOException
{
DataOutputBuffer out = new DataOutputBufferFixed(legacySerializedSize(withRemainingInPartition));
writeWithShortLength(null == partitionKey ? EMPTY_BYTE_BUFFER : partitionKey, out);
writeWithShortLength(null == rowMark ? EMPTY_BYTE_BUFFER : rowMark.mark, out);
out.writeInt(remaining);
if (withRemainingInPartition)
out.writeInt(remainingInPartition);
return out.buffer(false);
}
@VisibleForTesting
static boolean isLegacySerialized(ByteBuffer bytes)
{
int index = bytes.position();
int limit = bytes.limit();
if (limit - index < 2)
return false;
short partitionKeyLen = bytes.getShort(index);
if (partitionKeyLen < 0)
return false;
index += 2 + partitionKeyLen;
if (limit - index < 2)
return false;
short rowMarkerLen = bytes.getShort(index);
if (rowMarkerLen < 0)
return false;
index += 2 + rowMarkerLen;
if (limit - index < 4)
return false;
int remaining = bytes.getInt(index);
if (remaining < 0)
return false;
index += 4;
// V3 encoded by 2.1/2.2 - sans remainingInPartition
if (index == limit)
return true;
if (limit - index == 4)
{
int remainingInPartition = bytes.getInt(index);
return remainingInPartition >= 0; // the value must make sense
}
return false;
}
@SuppressWarnings({ "resource", "RedundantSuppression" })
private static PagingState legacyDeserialize(ByteBuffer bytes, ProtocolVersion protocolVersion) throws IOException
{
if (protocolVersion.isGreaterThan(ProtocolVersion.V3))
throw new IllegalArgumentException();
DataInputBuffer in = new DataInputBuffer(bytes, false);
ByteBuffer partitionKey = readWithShortLength(in);
ByteBuffer rawMark = readWithShortLength(in);
int remaining = in.readInt();
/*
* 2.1/2.2 implementations of V3 protocol did not write remainingInPartition, but C* 3.0+ does, so we need
* to handle both variants of V3 serialization for compatibility.
*/
int remainingInPartition = in.available() > 0 ? in.readInt() : Integer.MAX_VALUE;
return new PagingState(partitionKey.hasRemaining() ? partitionKey : null,
rawMark.hasRemaining() ? new RowMark(rawMark, protocolVersion) : null,
remaining,
remainingInPartition);
}
@VisibleForTesting
int legacySerializedSize(boolean withRemainingInPartition)
{
return serializedSizeWithShortLength(null == partitionKey ? EMPTY_BYTE_BUFFER : partitionKey)
+ serializedSizeWithShortLength(null == rowMark ? EMPTY_BYTE_BUFFER : rowMark.mark)
+ sizeof(remaining)
+ (withRemainingInPartition ? sizeof(remainingInPartition) : 0);
}
@Override
public final int hashCode()
{
return Objects.hash(partitionKey, rowMark, remaining, remainingInPartition);
}
@Override
public final boolean equals(Object o)
{
if(!(o instanceof PagingState))
return false;
PagingState that = (PagingState)o;
return Objects.equals(this.partitionKey, that.partitionKey)
&& Objects.equals(this.rowMark, that.rowMark)
&& this.remaining == that.remaining
&& this.remainingInPartition == that.remainingInPartition;
}
@Override
public String toString()
{
return String.format("PagingState(key=%s, cellname=%s, remaining=%d, remainingInPartition=%d",
partitionKey != null ? bytesToHex(partitionKey) : null,
rowMark,
remaining,
remainingInPartition);
}
/**
* Marks the last row returned by paging, the one from which paging should continue.
* This class essentially holds a row clustering, but due to backward compatibility reasons,
* we need to actually store the cell name for the last cell of the row we're marking when
* the protocol v3 is in use, and this class abstract that complication.
*
* See CASSANDRA-10254 for more details.
*/
public static class RowMark
{
// This can be null for convenience if no row is marked.
private final ByteBuffer mark;
private final ProtocolVersion protocolVersion;
private RowMark(ByteBuffer mark, ProtocolVersion protocolVersion)
{
this.mark = mark;
this.protocolVersion = protocolVersion;
}
private static List> makeClusteringTypes(TableMetadata metadata)
{
// This is the types that will be used when serializing the clustering in the paging state. We can't really use the actual clustering
// types however because we can't guarantee that there won't be a schema change between when we send the paging state and get it back,
// and said schema change could theoretically change one of the clustering types from a fixed width type to a non-fixed one
// (say timestamp -> blob). So we simply use a list of BytesTypes (for both reading and writting), which may be slightly inefficient
// for fixed-width types, but avoid any risk during schema changes.
int size = metadata.clusteringColumns().size();
List> l = new ArrayList<>(size);
for (int i = 0; i < size; i++)
l.add(BytesType.instance);
return l;
}
public static RowMark create(TableMetadata metadata, Row row, ProtocolVersion protocolVersion)
{
ByteBuffer mark;
if (protocolVersion.isSmallerOrEqualTo(ProtocolVersion.V3))
{
// We need to be backward compatible with 2.1/2.2 nodes paging states. Which means we have to send
// the full cellname of the "last" cell in the row we get (since that's how 2.1/2.2 nodes will start after
// that last row if they get that paging state).
Iterator> cells = row.cellsInLegacyOrder(metadata, true).iterator();
if (!cells.hasNext())
{
// If the last returned row has no cell, this means in 2.1/2.2 terms that we stopped on the row
// marker. Note that this shouldn't happen if the table is COMPACT STORAGE tables.
assert !metadata.isCompactTable();
mark = encodeCellName(metadata, row.clustering(), EMPTY_BYTE_BUFFER, null);
}
else
{
Cell cell = cells.next();
mark = encodeCellName(metadata, row.clustering(), cell.column().name.bytes, cell.column().isComplex() ? cell.path().get(0) : null);
}
}
else
{
// We froze the serialization version to 3.0 as we need to make this this doesn't change (that is, it has to be
// fix for a given version of the protocol).
mark = Clustering.serializer.serialize(row.clustering(), MessagingService.VERSION_30, makeClusteringTypes(metadata));
}
return new RowMark(mark, protocolVersion);
}
public Clustering clustering(TableMetadata metadata)
{
if (mark == null)
return null;
return protocolVersion.isSmallerOrEqualTo(ProtocolVersion.V3)
? decodeClustering(metadata, mark)
: Clustering.serializer.deserialize(mark, MessagingService.VERSION_30, makeClusteringTypes(metadata));
}
// Old (pre-3.0) encoding of cells. We need that for the protocol v3 as that is how things where encoded
private static ByteBuffer encodeCellName(TableMetadata metadata, Clustering clustering, ByteBuffer columnName, ByteBuffer collectionElement)
{
// v30 and v3X don't use composites for single-element clusterings in compact tables
if (metadata.isCompactTable() && metadata.comparator.size() == 1)
return clustering.bufferAt(0);
boolean isStatic = clustering == Clustering.STATIC_CLUSTERING;
// We use comparator.size() rather than clustering.size() because of static clusterings
int clusteringSize = metadata.comparator.size();
int size = clusteringSize + 1 + (collectionElement == null ? 0 : 1);
ByteBuffer[] values = new ByteBuffer[size];
for (int i = 0; i < clusteringSize; i++)
{
if (isStatic)
{
values[i] = EMPTY_BYTE_BUFFER;
continue;
}
ByteBuffer v = clustering.bufferAt(i);
// we can have null (only for dense compound tables for backward compatibility reasons) but that
// means we're done and should stop there as far as building the composite is concerned.
if (v == null)
return CompositeType.build(ByteBufferAccessor.instance, Arrays.copyOfRange(values, 0, i));
values[i] = v;
}
values[clusteringSize] = columnName;
if (collectionElement != null)
values[clusteringSize + 1] = collectionElement;
return CompositeType.build(ByteBufferAccessor.instance, isStatic, values);
}
private static Clustering decodeClustering(TableMetadata metadata, ByteBuffer value)
{
int csize = metadata.comparator.size();
if (csize == 0)
return Clustering.EMPTY;
// v30 and v3X don't use composites for single-element clusterings in compact tables
if (metadata.isCompactTable() && metadata.comparator.size() == 1)
return Clustering.make(value);
if (CompositeType.isStaticName(value, ByteBufferAccessor.instance))
return Clustering.STATIC_CLUSTERING;
List components = CompositeType.splitName(value, ByteBufferAccessor.instance);
return Clustering.make(components.subList(0, Math.min(csize, components.size())).toArray(new ByteBuffer[csize]));
}
@Override
public final int hashCode()
{
return Objects.hash(mark, protocolVersion);
}
@Override
public final boolean equals(Object o)
{
if(!(o instanceof RowMark))
return false;
RowMark that = (RowMark)o;
return Objects.equals(this.mark, that.mark) && this.protocolVersion == that.protocolVersion;
}
@Override
public String toString()
{
return mark == null ? "null" : bytesToHex(mark);
}
}
}
© 2015 - 2024 Weber Informatics LLC | Privacy Policy