com.unboundid.ldap.sdk.unboundidds.tools.SplitLDIFRDNHashTranslator Maven / Gradle / Ivy
/*
* Copyright 2016-2017 UnboundID Corp.
* All Rights Reserved.
*/
/*
* Copyright (C) 2016-2017 UnboundID Corp.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License (GPLv2 only)
* or the terms of the GNU Lesser General Public License (LGPLv2.1 only)
* as published by the Free Software Foundation.
*
* 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, see .
*/
package com.unboundid.ldap.sdk.unboundidds.tools;
import java.util.Collections;
import java.util.LinkedHashMap;
import java.util.LinkedHashSet;
import java.util.Map;
import java.util.Set;
import com.unboundid.ldap.sdk.DN;
import com.unboundid.ldap.sdk.Entry;
import com.unboundid.ldap.sdk.LDAPException;
import com.unboundid.ldif.LDIFException;
import com.unboundid.util.Debug;
import com.unboundid.util.ThreadSafety;
import com.unboundid.util.ThreadSafetyLevel;
import static com.unboundid.ldap.sdk.unboundidds.tools.ToolMessages.*;
/**
* This class provides an implementation of an LDIF reader entry translator that
* can be used to determine the set into which an entry should be placed by
* computing a modulus from a digest of the RDN.
*
*
* NOTE: This class is part of the Commercial Edition of the UnboundID
* LDAP SDK for Java. It is not available for use in applications that
* include only the Standard Edition of the LDAP SDK, and is not supported for
* use in conjunction with non-UnboundID products.
*
*/
@ThreadSafety(level=ThreadSafetyLevel.NOT_THREADSAFE)
final class SplitLDIFRDNHashTranslator
extends SplitLDIFTranslator
{
// A map of the names that will be used for each of the sets.
private final Map> setNames;
// The sets in which entries outside the split base should be placed.
private final Set outsideSplitBaseSetNames;
// The sets in which the split base entry should be placed.
private final Set splitBaseEntrySetNames;
/**
* Creates a new instance of this translator with the provided information.
*
* @param splitBaseDN The base DN below which to
* split entries.
* @param numSets The number of sets into which
* entries should be split.
* @param addEntriesOutsideSplitToAllSets Indicates whether entries
* outside the split should be
* added to all sets.
* @param addEntriesOutsideSplitToDedicatedSet Indicates whether entries
* outside the split should be
* added to all sets.
*/
SplitLDIFRDNHashTranslator(final DN splitBaseDN, final int numSets,
final boolean addEntriesOutsideSplitToAllSets,
final boolean addEntriesOutsideSplitToDedicatedSet)
{
super(splitBaseDN);
outsideSplitBaseSetNames = new LinkedHashSet(numSets+1);
splitBaseEntrySetNames = new LinkedHashSet(numSets);
if (addEntriesOutsideSplitToDedicatedSet)
{
outsideSplitBaseSetNames.add(SplitLDIFEntry.SET_NAME_OUTSIDE_SPLIT);
}
setNames = new LinkedHashMap>(numSets);
for (int i=0; i < numSets; i++)
{
final String setName = ".set" + (i+1);
setNames.put(i, Collections.singleton(setName));
splitBaseEntrySetNames.add(setName);
if (addEntriesOutsideSplitToAllSets)
{
outsideSplitBaseSetNames.add(setName);
}
}
}
/**
* {@inheritDoc}
*/
@Override()
public SplitLDIFEntry translate(final Entry original,
final long firstLineNumber)
throws LDIFException
{
// Get the parsed DN for the entry. If we can't, that's an error and we
// should only include it in the error set.
final DN dn;
try
{
dn = original.getParsedDN();
}
catch (final LDAPException le)
{
Debug.debugException(le);
return createEntry(original,
ERR_SPLIT_LDIF_RDN_HASH_TRANSLATOR_CANNOT_PARSE_DN.get(
le.getMessage()),
getErrorSetNames());
}
// If the parsed DN is outside the split base DN, then return the
// appropriate sets for that.
if (! dn.isDescendantOf(getSplitBaseDN(), true))
{
return createEntry(original, outsideSplitBaseSetNames);
}
// If the parsed DN matches the split base DN, then it will always go into
// all of the split sets. Otherwise, figure out where to send it based on
// the RDN component immediately below the split base DN.
if (dn.equals(getSplitBaseDN()))
{
return createEntry(original, splitBaseEntrySetNames);
}
else
{
return createFromRDNHash(original, dn, setNames);
}
}
}