/*
* 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.index;
import java.nio.ByteBuffer;
import java.nio.charset.CharacterCodingException;
import java.util.ArrayList;
import java.util.Collection;
import java.util.Collections;
import java.util.HashMap;
import java.util.HashSet;
import java.util.IdentityHashMap;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.ConcurrentMap;
import java.util.concurrent.ConcurrentNavigableMap;
import java.util.concurrent.ConcurrentSkipListMap;
import java.util.concurrent.Future;
import org.apache.cassandra.utils.ByteBufferUtil;
import org.apache.commons.lang3.StringUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.apache.cassandra.config.ColumnDefinition;
import org.apache.cassandra.config.IndexType;
import org.apache.cassandra.db.Cell;
import org.apache.cassandra.db.ColumnFamily;
import org.apache.cassandra.db.ColumnFamilyStore;
import org.apache.cassandra.db.DecoratedKey;
import org.apache.cassandra.db.IndexExpression;
import org.apache.cassandra.db.Row;
import org.apache.cassandra.db.SystemKeyspace;
import org.apache.cassandra.db.compaction.CompactionManager;
import org.apache.cassandra.db.composites.CellName;
import org.apache.cassandra.db.filter.ExtendedFilter;
import org.apache.cassandra.exceptions.ConfigurationException;
import org.apache.cassandra.exceptions.InvalidRequestException;
import org.apache.cassandra.io.sstable.ReducingKeyIterator;
import org.apache.cassandra.io.sstable.SSTableReader;
import org.apache.cassandra.utils.FBUtilities;
import org.apache.cassandra.utils.concurrent.OpOrder;
/**
* Manages all the indexes associated with a given CFS
* Different types of indexes can be created across the same CF
*/
public class SecondaryIndexManager
{
private static final Logger logger = LoggerFactory.getLogger(SecondaryIndexManager.class);
public static final Updater nullUpdater = new Updater()
{
public void insert(Cell cell) { }
public void update(Cell oldCell, Cell cell) { }
public void remove(Cell current) { }
public void updateRowLevelIndexes() {}
};
/**
* Organizes the indexes by column name
*/
private final ConcurrentNavigableMap<ByteBuffer, SecondaryIndex> indexesByColumn;
/**
* Keeps a single instance of a SecondaryIndex for many columns when the index type
* has isRowLevelIndex() == true
*
* This allows updates to happen to an entire row at once
*/
private final ConcurrentMap<Class<? extends SecondaryIndex>, SecondaryIndex> rowLevelIndexMap;
/**
* Keeps all secondary index instances, either per-column or per-row
*/
private final Set<SecondaryIndex> allIndexes;
/**
* The underlying column family containing the source data for these indexes
*/
public final ColumnFamilyStore baseCfs;
public SecondaryIndexManager(ColumnFamilyStore baseCfs)
{
indexesByColumn = new ConcurrentSkipListMap<>();
rowLevelIndexMap = new ConcurrentHashMap<>();
allIndexes = Collections.newSetFromMap(new ConcurrentHashMap<SecondaryIndex, Boolean>());
this.baseCfs = baseCfs;
}
/**
* Drops and adds new indexes associated with the underlying CF
*/
public void reload()
{
// figure out what needs to be added and dropped.
// future: if/when we have modifiable settings for secondary indexes,
// they'll need to be handled here.
Collection<ByteBuffer> indexedColumnNames = indexesByColumn.keySet();
for (ByteBuffer indexedColumn : indexedColumnNames)
{
ColumnDefinition def = baseCfs.metadata.getColumnDefinition(indexedColumn);
if (def == null || def.getIndexType() == null)
removeIndexedColumn(indexedColumn);
}
// TODO: allow all ColumnDefinition type
for (ColumnDefinition cdef : baseCfs.metadata.allColumns())
if (cdef.getIndexType() != null && !indexedColumnNames.contains(cdef.name.bytes))
addIndexedColumn(cdef);
for (SecondaryIndex index : allIndexes)
index.reload();
}
public Set<String> allIndexesNames()
{
Set<String> names = new HashSet<>(allIndexes.size());
for (SecondaryIndex index : allIndexes)
names.add(index.getIndexName());
return names;
}
/**
* Does a full, blocking rebuild of the indexes specified by columns from the sstables.
* Does nothing if columns is empty.
*
* Caller must acquire and release references to the sstables used here.
*
* @param sstables the data to build from
* @param idxNames the list of columns to index, ordered by comparator
*/
public void maybeBuildSecondaryIndexes(Collection<SSTableReader> sstables, Set<String> idxNames)
{
if (idxNames.isEmpty())
return;
logger.info(String.format("Submitting index build of %s for data in %s",
idxNames, StringUtils.join(sstables, ", ")));
SecondaryIndexBuilder builder = new SecondaryIndexBuilder(baseCfs, idxNames, new ReducingKeyIterator(sstables));
Future<?> future = CompactionManager.instance.submitIndexBuild(builder);
FBUtilities.waitOnFuture(future);
flushIndexesBlocking();
logger.info("Index build of {} complete", idxNames);
}
public boolean indexes(CellName name, Set<SecondaryIndex> indexes)
{
boolean matching = false;
for (SecondaryIndex index : indexes)
{
if (index.indexes(name))
{
matching = true;
break;
}
}
return matching;
}
public Set<SecondaryIndex> indexFor(CellName name, Set<SecondaryIndex> indexes)
{
Set<SecondaryIndex> matching = null;
for (SecondaryIndex index : indexes)
{
if (index.indexes(name))
{
if (matching == null)
matching = new HashSet<>();
matching.add(index);
}
}
return matching == null ? Collections.<SecondaryIndex>emptySet() : matching;
}
public boolean indexes(Cell cell)
{
return indexes(cell.name());
}
public boolean indexes(CellName name)
{
return indexes(name, allIndexes);
}
public Set<SecondaryIndex> indexFor(CellName name)
{
return indexFor(name, allIndexes);
}
/**
* @return true if at least one of the indexes can handle the clause.
*/
public boolean hasIndexFor(List<IndexExpression> clause)
{
if (clause == null || clause.isEmpty())
return false;
for (SecondaryIndexSearcher searcher : getIndexSearchersForQuery(clause))
if (searcher.canHandleIndexClause(clause))
return true;
return false;
}
/**
* Removes a existing index
* @param column the indexed column to remove
*/
public void removeIndexedColumn(ByteBuffer column)
{
SecondaryIndex index = indexesByColumn.remove(column);
if (index == null)
return;
// Remove this column from from row level index map as well as all indexes set
if (index instanceof PerRowSecondaryIndex)
{
index.removeColumnDef(column);
// If no columns left remove from row level lookup as well as all indexes set
if (index.getColumnDefs().isEmpty())
{
allIndexes.remove(index);
rowLevelIndexMap.remove(index.getClass());
}
}
else
{
allIndexes.remove(index);
}
index.removeIndex(column);
SystemKeyspace.setIndexRemoved(baseCfs.metadata.ksName, index.getNameForSystemKeyspace(column));
}
/**
* Adds and builds a index for a column
* @param cdef the column definition holding the index data
* @return a future which the caller can optionally block on signaling the index is built
*/
public synchronized Future<?> addIndexedColumn(ColumnDefinition cdef)
{
if (indexesByColumn.containsKey(cdef.name.bytes))
return null;
assert cdef.getIndexType() != null;
SecondaryIndex index;
try
{
index = SecondaryIndex.createInstance(baseCfs, cdef);
}
catch (ConfigurationException e)
{
throw new RuntimeException(e);
}
// Keep a single instance of the index per-cf for row level indexes
// since we want all columns to be under the index
if (index instanceof PerRowSecondaryIndex)
{
SecondaryIndex currentIndex = rowLevelIndexMap.get(index.getClass());
if (currentIndex == null)
{
rowLevelIndexMap.put(index.getClass(), index);
index.init();
}
else
{
index = currentIndex;
index.addColumnDef(cdef);
logger.info("Creating new index : {}",cdef);
}
}
else
{
// TODO: We sould do better than throw a RuntimeException
if (cdef.getIndexType() == IndexType.CUSTOM && index instanceof AbstractSimplePerColumnSecondaryIndex)
throw new RuntimeException("Cannot use a subclass of AbstractSimplePerColumnSecondaryIndex as a CUSTOM index, as they assume they are CFS backed");
index.init();
}
// link in indexedColumns. this means that writes will add new data to
// the index immediately,
// so we don't have to lock everything while we do the build. it's up to
// the operator to wait
// until the index is actually built before using in queries.
indexesByColumn.put(cdef.name.bytes, index);
// Add to all indexes set:
allIndexes.add(index);
// if we're just linking in the index to indexedColumns on an
// already-built index post-restart, we're done
if (index.isIndexBuilt(cdef.name.bytes))
return null;
return index.buildIndexAsync();
}
/**
*
* @param column the name of indexes column
* @return the index
*/
public SecondaryIndex getIndexForColumn(ByteBuffer column)
{
return indexesByColumn.get(column);
}
/**
* Remove the index
*/
public void invalidate()
{
for (SecondaryIndex index : allIndexes)
index.invalidate();
}
/**
* Flush all indexes to disk
*/
public void flushIndexesBlocking()
{
// despatch flushes for all CFS backed indexes
List<Future<?>> wait = new ArrayList<>();
synchronized (baseCfs.getDataTracker())
{
for (SecondaryIndex index : allIndexes)
if (index.getIndexCfs() != null)
wait.add(index.getIndexCfs().forceFlush());
}
// blockingFlush any non-CFS-backed indexes
for (SecondaryIndex index : allIndexes)
if (index.getIndexCfs() == null)
index.forceBlockingFlush();
// wait for the CFS-backed index flushes to complete
FBUtilities.waitOnFutures(wait);
}
/**
* @return all built indexes (ready to use)
*/
public List<String> getBuiltIndexes()
{
List<String> indexList = new ArrayList<>();
for (Map.Entry<ByteBuffer, SecondaryIndex> entry : indexesByColumn.entrySet())
{
SecondaryIndex index = entry.getValue();
if (index.isIndexBuilt(entry.getKey()))
indexList.add(entry.getValue().getIndexName());
}
return indexList;
}
/**
* @return all CFS from indexes which use a backing CFS internally (KEYS)
*/
public Set<ColumnFamilyStore> getIndexesBackedByCfs()
{
Set<ColumnFamilyStore> cfsList = new HashSet<>();
for (SecondaryIndex index: allIndexes)
{
ColumnFamilyStore cfs = index.getIndexCfs();
if (cfs != null)
cfsList.add(cfs);
}
return cfsList;
}
/**
* @return all indexes which do *not* use a backing CFS internally
*/
public Set<SecondaryIndex> getIndexesNotBackedByCfs()
{
// we use identity map because per row indexes use same instance across many columns
Set<SecondaryIndex> indexes = Collections.newSetFromMap(new IdentityHashMap<SecondaryIndex, Boolean>());
for (SecondaryIndex index: allIndexes)
if (index.getIndexCfs() == null)
indexes.add(index);
return indexes;
}
/**
* @return all of the secondary indexes without distinction to the (non-)backed by secondary ColumnFamilyStore.
*/
public Set<SecondaryIndex> getIndexes()
{
return allIndexes;
}
/**
* @return if there are ANY indexes for this table..
*/
public boolean hasIndexes()
{
return !indexesByColumn.isEmpty();
}
/**
* When building an index against existing data, add the given row to the index
*
* @param key the row key
* @param cf the current rows data
*/
public void indexRow(ByteBuffer key, ColumnFamily cf, OpOrder.Group opGroup)
{
// Update entire row only once per row level index
Set<Class<? extends SecondaryIndex>> appliedRowLevelIndexes = null;
for (SecondaryIndex index : allIndexes)
{
if (index instanceof PerRowSecondaryIndex)
{
if (appliedRowLevelIndexes == null)
appliedRowLevelIndexes = new HashSet<>();
if (appliedRowLevelIndexes.add(index.getClass()))
((PerRowSecondaryIndex)index).index(key, cf);
}
else
{
for (Cell cell : cf)
if (cell.isLive() && index.indexes(cell.name()))
((PerColumnSecondaryIndex) index).insert(key, cell, opGroup);
}
}
}
/**
* Delete all columns from all indexes for this row. For when cleanup rips a row out entirely.
*
* @param key the row key
* @param indexedColumnsInRow all column names in row
*/
public void deleteFromIndexes(DecoratedKey key, List<Cell> indexedColumnsInRow, OpOrder.Group opGroup)
{
// Update entire row only once per row level index
Set<Class<? extends SecondaryIndex>> cleanedRowLevelIndexes = null;
for (Cell cell : indexedColumnsInRow)
{
for (SecondaryIndex index : indexFor(cell.name()))
{
if (index instanceof PerRowSecondaryIndex)
{
if (cleanedRowLevelIndexes == null)
cleanedRowLevelIndexes = new HashSet<>();
if (cleanedRowLevelIndexes.add(index.getClass()))
((PerRowSecondaryIndex) index).delete(key, opGroup);
}
else
{
((PerColumnSecondaryIndex) index).deleteForCleanup(key.getKey(), cell, opGroup);
}
}
}
}
/**
* This helper acts as a closure around the indexManager
* and updated cf data to ensure that down in
* Memtable's ColumnFamily implementation, the index
* can get updated. Note: only a CF backed by AtomicSortedColumns implements
* this behaviour fully, other types simply ignore the index updater.
*/
public Updater updaterFor(DecoratedKey key, ColumnFamily cf, OpOrder.Group opGroup)
{
return (indexesByColumn.isEmpty() && rowLevelIndexMap.isEmpty())
? nullUpdater
: new StandardUpdater(key, cf, opGroup);
}
/**
* Updated closure with only the modified row key.
*/
public Updater gcUpdaterFor(DecoratedKey key)
{
return new GCUpdater(key);
}
/**
* Get a list of IndexSearchers from the union of expression index types
* @param clause the query clause
* @return the searchers needed to query the index
*/
public List<SecondaryIndexSearcher> getIndexSearchersForQuery(List<IndexExpression> clause)
{
Map<String, Set<ByteBuffer>> groupByIndexType = new HashMap<>();
//Group columns by type
for (IndexExpression ix : clause)
{
SecondaryIndex index = getIndexForColumn(ix.column);
if (index == null || !index.supportsOperator(ix.operator))
continue;
Set<ByteBuffer> columns = groupByIndexType.get(index.indexTypeForGrouping());
if (columns == null)
{
columns = new HashSet<>();
groupByIndexType.put(index.indexTypeForGrouping(), columns);
}
columns.add(ix.column);
}
List<SecondaryIndexSearcher> indexSearchers = new ArrayList<>(groupByIndexType.size());
//create searcher per type
for (Set<ByteBuffer> column : groupByIndexType.values())
indexSearchers.add(getIndexForColumn(column.iterator().next()).createSecondaryIndexSearcher(column));
return indexSearchers;
}
/**
* Validates an union of expression index types. It will throw a {@link RuntimeException} if
* any of the expressions in the provided clause is not valid for its index implementation.
* @param clause the query clause
* @throws org.apache.cassandra.exceptions.InvalidRequestException in case of validation errors
*/
public void validateIndexSearchersForQuery(List<IndexExpression> clause) throws InvalidRequestException
{
// Group by index type
Map<String, Set<IndexExpression>> expressionsByIndexType = new HashMap<>();
Map<String, Set<ByteBuffer>> columnsByIndexType = new HashMap<>();
for (IndexExpression indexExpression : clause)
{
SecondaryIndex index = getIndexForColumn(indexExpression.column);
if (index == null)
continue;
String canonicalIndexName = index.getClass().getCanonicalName();
Set<IndexExpression> expressions = expressionsByIndexType.get(canonicalIndexName);
Set<ByteBuffer> columns = columnsByIndexType.get(canonicalIndexName);
if (expressions == null)
{
expressions = new HashSet<>();
columns = new HashSet<>();
expressionsByIndexType.put(canonicalIndexName, expressions);
columnsByIndexType.put(canonicalIndexName, columns);
}
expressions.add(indexExpression);
columns.add(indexExpression.column);
}
// Validate
boolean haveSupportedIndexLookup = false;
for (Map.Entry<String, Set<IndexExpression>> expressions : expressionsByIndexType.entrySet())
{
Set<ByteBuffer> columns = columnsByIndexType.get(expressions.getKey());
SecondaryIndex secondaryIndex = getIndexForColumn(columns.iterator().next());
SecondaryIndexSearcher searcher = secondaryIndex.createSecondaryIndexSearcher(columns);
for (IndexExpression expression : expressions.getValue())
{
searcher.validate(expression);
haveSupportedIndexLookup |= secondaryIndex.supportsOperator(expression.operator);
}
}
if (!haveSupportedIndexLookup)
{
// build the error message
int i = 0;
StringBuilder sb = new StringBuilder("No secondary indexes on the restricted columns support the provided operators: ");
for (Map.Entry<String, Set<IndexExpression>> expressions : expressionsByIndexType.entrySet())
{
for (IndexExpression expression : expressions.getValue())
{
if (i++ > 0)
sb.append(", ");
sb.append("'");
String columnName;
try
{
columnName = ByteBufferUtil.string(expression.column);
}
catch (CharacterCodingException ex)
{
columnName = "<unprintable>";
}
sb.append(columnName).append(" ").append(expression.operator).append(" <value>").append("'");
}
}
throw new InvalidRequestException(sb.toString());
}
}
/**
* Performs a search across a number of column indexes
*
* @param filter the column range to restrict to
* @return found indexed rows
*/
public List<Row> search(ExtendedFilter filter)
{
SecondaryIndexSearcher mostSelective = getHighestSelectivityIndexSearcher(filter.getClause());
if (mostSelective == null)
return Collections.emptyList();
else
return mostSelective.search(filter);
}
public Set<SecondaryIndex> getIndexesByNames(Set<String> idxNames)
{
Set<SecondaryIndex> result = new HashSet<>();
for (SecondaryIndex index : allIndexes)
if (idxNames.contains(index.getIndexName()))
result.add(index);
return result;
}
public void setIndexBuilt(Set<String> idxNames)
{
for (SecondaryIndex index : getIndexesByNames(idxNames))
index.setIndexBuilt();
}
public void setIndexRemoved(Set<String> idxNames)
{
for (SecondaryIndex index : getIndexesByNames(idxNames))
index.setIndexRemoved();
}
public boolean validate(Cell cell)
{
for (SecondaryIndex index : indexFor(cell.name()))
{
if (!index.validate(cell))
return false;
}
return true;
}
static boolean shouldCleanupOldValue(Cell oldCell, Cell newCell)
{
// If any one of name/value/timestamp are different, then we
// should delete from the index. If not, then we can infer that
// at least one of the cells is an ExpiringColumn and that the
// difference is in the expiry time. In this case, we don't want to
// delete the old value from the index as the tombstone we insert
// will just hide the inserted value.
// Completely identical cells (including expiring columns with
// identical ttl & localExpirationTime) will not get this far due
// to the oldCell.equals(newColumn) in StandardUpdater.update
return !oldCell.name().equals(newCell.name())
|| !oldCell.value().equals(newCell.value())
|| oldCell.timestamp() != newCell.timestamp();
}
public static interface Updater
{
/** called when constructing the index against pre-existing data */
public void insert(Cell cell);
/** called when updating the index from a memtable */
public void update(Cell oldCell, Cell cell);
/** called when lazy-updating the index during compaction (CASSANDRA-2897) */
public void remove(Cell current);
/** called after memtable updates are complete (CASSANDRA-5397) */
public void updateRowLevelIndexes();
}
private final class GCUpdater implements Updater
{
private final DecoratedKey key;
public GCUpdater(DecoratedKey key)
{
this.key = key;
}
public void insert(Cell cell)
{
throw new UnsupportedOperationException();
}
public void update(Cell oldCell, Cell newCell)
{
throw new UnsupportedOperationException();
}
public void remove(Cell cell)
{
if (!cell.isLive())
return;
for (SecondaryIndex index : indexFor(cell.name()))
{
if (index instanceof PerColumnSecondaryIndex)
{
try (OpOrder.Group opGroup = baseCfs.keyspace.writeOrder.start())
{
((PerColumnSecondaryIndex) index).delete(key.getKey(), cell, opGroup);
}
}
}
}
public void updateRowLevelIndexes()
{
for (SecondaryIndex index : rowLevelIndexMap.values())
((PerRowSecondaryIndex) index).index(key.getKey(), null);
}
}
private final class StandardUpdater implements Updater
{
private final DecoratedKey key;
private final ColumnFamily cf;
private final OpOrder.Group opGroup;
public StandardUpdater(DecoratedKey key, ColumnFamily cf, OpOrder.Group opGroup)
{
this.key = key;
this.cf = cf;
this.opGroup = opGroup;
}
public void insert(Cell cell)
{
if (!cell.isLive())
return;
for (SecondaryIndex index : indexFor(cell.name()))
if (index instanceof PerColumnSecondaryIndex)
((PerColumnSecondaryIndex) index).insert(key.getKey(), cell, opGroup);
}
public void update(Cell oldCell, Cell cell)
{
if (oldCell.equals(cell))
return;
for (SecondaryIndex index : indexFor(cell.name()))
{
if (index instanceof PerColumnSecondaryIndex)
{
if (cell.isLive())
{
((PerColumnSecondaryIndex) index).update(key.getKey(), oldCell, cell, opGroup);
}
else
{
// Usually we want to delete the old value from the index, except when
// name/value/timestamp are all equal, but the columns themselves
// are not (as is the case when overwriting expiring columns with
// identical values and ttl) Then, we don't want to delete as the
// tombstone will hide the new value we just inserted; see CASSANDRA-7268
if (shouldCleanupOldValue(oldCell, cell))
((PerColumnSecondaryIndex) index).delete(key.getKey(), oldCell, opGroup);
}
}
}
}
public void remove(Cell cell)
{
if (!cell.isLive())
return;
for (SecondaryIndex index : indexFor(cell.name()))
if (index instanceof PerColumnSecondaryIndex)
((PerColumnSecondaryIndex) index).delete(key.getKey(), cell, opGroup);
}
public void updateRowLevelIndexes()
{
for (SecondaryIndex index : rowLevelIndexMap.values())
((PerRowSecondaryIndex) index).index(key.getKey(), cf);
}
}
public SecondaryIndexSearcher getHighestSelectivityIndexSearcher(List<IndexExpression> clause)
{
List<SecondaryIndexSearcher> indexSearchers = getIndexSearchersForQuery(clause);
if (indexSearchers.isEmpty())
return null;
SecondaryIndexSearcher mostSelective = null;
long bestEstimate = Long.MAX_VALUE;
for (SecondaryIndexSearcher searcher : indexSearchers)
{
SecondaryIndex highestSelectivityIndex = searcher.highestSelectivityIndex(clause);
long estimate = highestSelectivityIndex.estimateResultRows();
if (estimate <= bestEstimate)
{
bestEstimate = estimate;
mostSelective = searcher;
}
}
return mostSelective;
}
}