/**
Copyright (C) SYSTAP, LLC DBA Blazegraph 2006-2016. All rights reserved.
Contact:
SYSTAP, LLC DBA Blazegraph
2501 Calvert ST NW #106
Washington, DC 20008
licenses@blazegraph.com
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
/*
* Created on May 21, 2007
*/
package com.bigdata.rdf.lexicon;
import com.bigdata.btree.IIndex;
import com.bigdata.btree.proc.AbstractKeyArrayIndexProcedure;
import com.bigdata.btree.proc.AbstractKeyArrayIndexProcedureConstructor;
import com.bigdata.btree.proc.IParallelizableIndexProcedure;
import com.bigdata.btree.proc.IResultHandler;
import com.bigdata.btree.raba.IRaba;
import com.bigdata.btree.raba.codec.IRabaCoder;
import com.bigdata.rdf.internal.impl.TermId;
import com.bigdata.rdf.model.BigdataValueSerializer;
import com.bigdata.relation.IMutableRelationIndexWriteProcedure;
import com.bigdata.service.ndx.NopAggregator;
/**
* Unisolated write operation makes consistent assertions on the
* <em>id:term</em> index based on the data developed by the {@link Term2IdWriteProc}
* operation.
*
* @author <a href="mailto:thompsonbry@users.sourceforge.net">Bryan Thompson</a>
*/
public class Id2TermWriteProc extends AbstractKeyArrayIndexProcedure<Void> implements
IParallelizableIndexProcedure<Void>, IMutableRelationIndexWriteProcedure<Void> {
/**
*
*/
private static final long serialVersionUID = -5480378815444534653L;
/**
* Enables validation that a pre-assigned term identifier is being
* consistently mapped onto the same term. Errors are reported if, for
* example, the index has a record that a term identifier is mapped onto one
* URL but the procedure was invoked with a different URI paired to that
* term identifiers. When such errors are reported, they generally indicate
* a problem with the TERM2ID index where it is failing to maintain a
* consistent mapping.
* <p>
* Validation may be disabled for releases, however it is not really that
* much overhead since the operation is on the in-memory representation.
*
* @deprecated Validation can not be reasonably applied it the Unicode
* collation is less than Identical. It also has problems for
* datatype literals if different lexical forms are all mapped
* onto the same key,e.g.,
*
* <pre>
* 12ˆˆ<xsd:float>
* 12.0ˆˆ<xsd:float>
* 12.00ˆˆ<xsd:float>
* </pre>
*
* will all be mapped to the same key and hence would give the
* appearance of a conflict if we were to reject any of these
* forms when another of the forms was already present under the
* key.
*
* Note: Now it's not only deprecated, but the code that relies on it has
* been commented out. This is because it makes assumptions about how
* {@link TermId} objects are encoded and decoded. Under the legacy model
* they were simple longs. After the lexicon refactor we use the byte
* flags also. So if we ever decide to do validation again here, we need
* to figure out how to give this class access to an {@link IIVEncoder}.
*/
static private transient final boolean validate = false;
@Override
public final boolean isReadOnly() {
return false;
}
/**
* De-serialization constructor.
*/
public Id2TermWriteProc() {
}
protected Id2TermWriteProc(final IRabaCoder keysCoder, final IRabaCoder valsCoder,
int fromIndex, int toIndex, byte[][] keys, byte[][] vals) {
super(keysCoder, valsCoder, fromIndex, toIndex, keys, vals);
assert vals != null;
}
public static class Id2TermWriteProcConstructor extends
AbstractKeyArrayIndexProcedureConstructor<Id2TermWriteProc> {
public static Id2TermWriteProcConstructor INSTANCE = new Id2TermWriteProcConstructor();
/**
* Values are required.
*/
@Override
public final boolean sendValues() {
return true;
}
private Id2TermWriteProcConstructor() {}
@Override
public Id2TermWriteProc newInstance(final IRabaCoder keysCoder,
final IRabaCoder valsCoder, final int fromIndex,
final int toIndex, final byte[][] keys, final byte[][] vals) {
return new Id2TermWriteProc(keysCoder, valsCoder, fromIndex, toIndex,
keys, vals);
}
}
/**
* Conditionally inserts each key-value pair into the index. The keys are
* the term identifiers. The values are the terms as serialized by
* {@link BigdataValueSerializer}. Since a conditional insert is used, the
* operation does not cause terms that are already known to the ids index to
* be re-inserted, thereby reducing writes of dirty index nodes.
*
* @param ndx
* The index.
*
* @return <code>null</code>.
*/
@Override
public Void applyOnce(final IIndex ndx, final IRaba keys, final IRaba vals) {
final int n = keys.size();
for (int i = 0; i < n; i++) {
// Note: the key is the term identifier.
// @todo copy key/val into reused buffers to reduce allocation.
final byte[] key = keys.get(i);
// // Note: the value is the serialized term (and never a BNode).
// final byte[] val;
//
// if (validate) {
//
// // The term identifier.
// final long id = KeyBuilder.decodeLong(key, 0);
//
// assert id != TermId.NULL;
//
// // Note: BNodes are not allowed in the reverse index.
// assert ! VTE.isBNode(id);
//
// // Note: SIDS are not allowed in the reverse index.
// assert ! VTE.isStatement(id);
//
// /*
// * When the term identifier is found in the reverse mapping
// * this code path validates that the serialized term is the
// * same.
// */
// final byte[] oldval = ndx.lookup(key);
//
// val = getValue(i);
//
// if( oldval == null ) {
//
// if (ndx.insert(key, val) != null) {
//
// throw new AssertionError();
//
// }
//
// } else {
//
// /*
// * Note: This would fail if the serialization of the term
// * was changed for an existing database instance. In order
// * to validate when different serialization formats might be
// * in use you have to actually deserialize the terms.
// * However, I have the validation logic here just as a
// * sanity check while getting the basic system running - it
// * is not meant to be deployed.
// */
//
// if (! BytesUtil.bytesEqual(val, oldval)) {
//
// final char suffix;
// if (VTE.isLiteral(id))
// suffix = 'L';
// else if (VTE.isURI(id))
// suffix = 'U';
// else if (VTE.isBNode(id))
// suffix = 'B';
// else if (VTE.isStatement(id))
// suffix = 'S';
// else
// suffix = '?';
//
// /*
// * We have to go one step further and compare the
// * deserialized value in order to decide if there is
// * really an inconsistency in the index. For example,
// * "abc@en" and "abc@EN" encode as different byte[]s,
// * but they are EQUALS() for RDF since the language code
// * comparison is case insensitive. The same problem can
// * occur for data type literals, since lexically
// * distinct literals are are mapped onto the same point
// * in the data type space (the same key). However,
// * comparison based on data type equality is not really
// * provided for by BigdataLiteral, so we get into
// * trouble if we attempt to detect errors based on
// * datatype literals.
// */
// final BigdataValueSerializer valSer = new BigdataValueSerializer(
// new ValueFactoryImpl());
//
// final Value term = valSer.deserialize(val);
// final Value oldterm = valSer.deserialize(oldval);
//
// if (!term.equals(oldterm)) {
//
// log.error("term=" + term);
// log.error("oldterm=" + oldterm);
// log.error("id=" + id + suffix);
// log.error("key=" + BytesUtil.toString(key));
// log.error("val=" + Arrays.toString(val));
// log.error("oldval=" + Arrays.toString(oldval));
// if (ndx.getIndexMetadata().getPartitionMetadata() != null)
// log.error(ndx.getIndexMetadata()
// .getPartitionMetadata().toString());
//
// throw new RuntimeException(
// "Consistency problem: id=" + id);
// }
//
// }
//
// }
//
// } else {
/*
* This code path does not validate that the term identifier
* is mapped to the same term. This is the code path that
* you SHOULD use.
*/
// See BLZG-1539
ndx.putIfAbsent(key, vals.get(i));
// if (!ndx.contains(key)) {
//
// val = vals.get(i);
//
// if (ndx.insert(key, val) != null) {
//
// throw new AssertionError();
//
// }
//
// }
// }
}
return null;
}
/**
* Nothing is returned, so nothing to aggregate, but uses a
* {@link NopAggregator} to preserve striping against a local index.
*/
@SuppressWarnings("unchecked")
@Override
protected IResultHandler<Void, Void> newAggregator() {
// NOP aggegrator preserves striping against the index.
return NopAggregator.INSTANCE;
}
}