| 1 |
26600
|
sandro.lab
|
package eu.dnetlib.data.mapreduce.util;
|
| 2 |
|
|
|
| 3 |
|
|
import java.nio.ByteBuffer;
|
| 4 |
|
|
|
| 5 |
|
|
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
|
| 6 |
27941
|
claudio.at
|
import org.apache.hadoop.hbase.util.Bytes;
|
| 7 |
26600
|
sandro.lab
|
|
| 8 |
28058
|
claudio.at
|
import eu.dnetlib.data.proto.DedupProtos.Dedup;
|
| 9 |
|
|
import eu.dnetlib.data.proto.DedupSimilarityProtos.DedupSimilarity;
|
| 10 |
28308
|
claudio.at
|
import eu.dnetlib.data.proto.FieldTypeProtos.Qualifier;
|
| 11 |
|
|
import eu.dnetlib.data.proto.OafProtos.OafRel;
|
| 12 |
|
|
import eu.dnetlib.data.proto.OafProtos.OafRel.Builder;
|
| 13 |
|
|
import eu.dnetlib.data.proto.OrganizationOrganizationProtos.OrganizationOrganization;
|
| 14 |
|
|
import eu.dnetlib.data.proto.PersonPersonProtos.PersonPerson;
|
| 15 |
|
|
import eu.dnetlib.data.proto.RelMetadataProtos.RelMetadata;
|
| 16 |
27941
|
claudio.at
|
import eu.dnetlib.data.proto.RelTypeProtos.RelType;
|
| 17 |
|
|
import eu.dnetlib.data.proto.RelTypeProtos.SubRelType;
|
| 18 |
28308
|
claudio.at
|
import eu.dnetlib.data.proto.ResultResultProtos.ResultResult;
|
| 19 |
27941
|
claudio.at
|
import eu.dnetlib.data.proto.TypeProtos.Type;
|
| 20 |
28308
|
claudio.at
|
import eu.dnetlib.pace.util.DedupConfig;
|
| 21 |
26600
|
sandro.lab
|
|
| 22 |
27941
|
claudio.at
|
public class DedupUtils {
|
| 23 |
|
|
|
| 24 |
|
|
public static final String CF_SEPARATOR = "_";
|
| 25 |
|
|
|
| 26 |
26600
|
sandro.lab
|
public static final String ROOT = "dedup_wf";
|
| 27 |
|
|
|
| 28 |
28308
|
claudio.at
|
public static final String BODY_S = "body";
|
| 29 |
|
|
|
| 30 |
|
|
public static final byte[] BODY_B = Bytes.toBytes(BODY_S);
|
| 31 |
|
|
|
| 32 |
|
|
public static String dedupPrefix(final String dedupRun) {
|
| 33 |
26600
|
sandro.lab
|
return "|" + ROOT + "_" + dedupRun + "::";
|
| 34 |
|
|
}
|
| 35 |
27941
|
claudio.at
|
|
| 36 |
28308
|
claudio.at
|
public static String newId(final String id, final String dedupRun) {
|
| 37 |
|
|
if ((dedupRun == null) || (dedupRun.length() != 3)) throw new IllegalArgumentException("wrong dedupRun param");
|
| 38 |
27941
|
claudio.at
|
|
| 39 |
26600
|
sandro.lab
|
return id.replaceFirst("\\|.*\\:\\:", dedupPrefix(dedupRun));
|
| 40 |
|
|
}
|
| 41 |
|
|
|
| 42 |
28308
|
claudio.at
|
public static byte[] newIdBytes(final String s, final String dedupRun) {
|
| 43 |
26600
|
sandro.lab
|
return newId(s, dedupRun).getBytes();
|
| 44 |
|
|
}
|
| 45 |
|
|
|
| 46 |
28308
|
claudio.at
|
public static byte[] newIdBytes(final ByteBuffer b, final String dedupRun) {
|
| 47 |
26600
|
sandro.lab
|
return newId(new String(b.array()), dedupRun).getBytes();
|
| 48 |
|
|
}
|
| 49 |
|
|
|
| 50 |
28308
|
claudio.at
|
public static boolean isRoot(final String s) {
|
| 51 |
26600
|
sandro.lab
|
return s.contains(ROOT);
|
| 52 |
|
|
}
|
| 53 |
27941
|
claudio.at
|
|
| 54 |
28308
|
claudio.at
|
public static boolean isRoot(final ImmutableBytesWritable s) {
|
| 55 |
26600
|
sandro.lab
|
return isRoot(s.copyBytes());
|
| 56 |
|
|
}
|
| 57 |
27941
|
claudio.at
|
|
| 58 |
28308
|
claudio.at
|
public static boolean isRoot(final byte[] s) {
|
| 59 |
26600
|
sandro.lab
|
return isRoot(new String(s));
|
| 60 |
|
|
}
|
| 61 |
27941
|
claudio.at
|
|
| 62 |
28308
|
claudio.at
|
public static String getDedupCF_merges(final Type type) {
|
| 63 |
28058
|
claudio.at
|
return getRelType(type) + CF_SEPARATOR + SubRelType.dedup + CF_SEPARATOR + Dedup.RelName.merges;
|
| 64 |
27941
|
claudio.at
|
}
|
| 65 |
|
|
|
| 66 |
28308
|
claudio.at
|
public static byte[] getDedupCF_mergesBytes(final Type type) {
|
| 67 |
28058
|
claudio.at
|
return Bytes.toBytes(getDedupCF_merges(type));
|
| 68 |
27941
|
claudio.at
|
}
|
| 69 |
|
|
|
| 70 |
28308
|
claudio.at
|
public static String getDedupCF_mergedIn(final Type type) {
|
| 71 |
28058
|
claudio.at
|
return getRelType(type) + CF_SEPARATOR + SubRelType.dedup + CF_SEPARATOR + Dedup.RelName.isMergedIn;
|
| 72 |
|
|
}
|
| 73 |
|
|
|
| 74 |
28308
|
claudio.at
|
public static byte[] getDedupCF_mergedInBytes(final Type type) {
|
| 75 |
28058
|
claudio.at
|
return Bytes.toBytes(getDedupCF_mergedIn(type));
|
| 76 |
|
|
}
|
| 77 |
|
|
|
| 78 |
28308
|
claudio.at
|
public static String getSimilarityCF(final Type type) {
|
| 79 |
28058
|
claudio.at
|
return getRelType(type) + CF_SEPARATOR + SubRelType.dedupSimilarity + CF_SEPARATOR + DedupSimilarity.RelName.isSimilarTo;
|
| 80 |
27941
|
claudio.at
|
}
|
| 81 |
|
|
|
| 82 |
28308
|
claudio.at
|
public static byte[] getSimilarityCFBytes(final Type type) {
|
| 83 |
27941
|
claudio.at
|
return Bytes.toBytes(getSimilarityCF(type));
|
| 84 |
|
|
}
|
| 85 |
|
|
|
| 86 |
28308
|
claudio.at
|
public static String getRelTypeString(final Type type) {
|
| 87 |
27941
|
claudio.at
|
return getRelType(type).toString();
|
| 88 |
|
|
}
|
| 89 |
|
|
|
| 90 |
28308
|
claudio.at
|
public static RelType getRelType(final Type type) {
|
| 91 |
27941
|
claudio.at
|
switch (type) {
|
| 92 |
|
|
case organization:
|
| 93 |
|
|
return RelType.organizationOrganization;
|
| 94 |
|
|
case person:
|
| 95 |
|
|
return RelType.personPerson;
|
| 96 |
|
|
case result:
|
| 97 |
|
|
return RelType.resultResult;
|
| 98 |
|
|
default:
|
| 99 |
|
|
throw new IllegalArgumentException("Deduplication not supported for entity type: " + type);
|
| 100 |
|
|
}
|
| 101 |
|
|
}
|
| 102 |
|
|
|
| 103 |
28308
|
claudio.at
|
public static ColumnFamily decodeCF(final byte[] b) {
|
| 104 |
27941
|
claudio.at
|
String[] s = new String(b).split(CF_SEPARATOR);
|
| 105 |
|
|
return new DedupUtils().getCF(RelType.valueOf(s[0]), SubRelType.valueOf(s[1]));
|
| 106 |
|
|
}
|
| 107 |
|
|
|
| 108 |
28308
|
claudio.at
|
private ColumnFamily getCF(final RelType relType, final SubRelType subRelType) {
|
| 109 |
27941
|
claudio.at
|
return new ColumnFamily(relType, subRelType);
|
| 110 |
|
|
}
|
| 111 |
|
|
|
| 112 |
28308
|
claudio.at
|
public static OafRel.Builder getDedup(final DedupConfig dedupConf, final String from, final String to, final Dedup.RelName relClass) {
|
| 113 |
28411
|
claudio.at
|
Type type = Type.valueOf(dedupConf.getEntityType());
|
| 114 |
|
|
RelType relType = DedupUtils.getRelType(type);
|
| 115 |
28308
|
claudio.at
|
Builder oafRel =
|
| 116 |
|
|
OafRel.newBuilder().setRelType(relType).setSubRelType(SubRelType.dedup).setRelClass(relClass.toString()).setChild(false)
|
| 117 |
|
|
.setSource(new String(from)).setTarget(new String(to));
|
| 118 |
28411
|
claudio.at
|
switch (type) {
|
| 119 |
28308
|
claudio.at
|
case organization:
|
| 120 |
|
|
oafRel.setOrganizationOrganization(OrganizationOrganization.newBuilder().setDedup(
|
| 121 |
|
|
DedupUtils.dedup(relClass, "dnet:organization_organization_relations")));
|
| 122 |
|
|
break;
|
| 123 |
|
|
case person:
|
| 124 |
|
|
oafRel.setPersonPerson(PersonPerson.newBuilder().setDedup(DedupUtils.dedup(relClass, "dnet:person_person_relations")));
|
| 125 |
|
|
break;
|
| 126 |
|
|
case result:
|
| 127 |
|
|
oafRel.setResultResult(ResultResult.newBuilder().setDedup(DedupUtils.dedup(relClass, "dnet:result_result_relations")));
|
| 128 |
|
|
break;
|
| 129 |
|
|
default:
|
| 130 |
|
|
throw new IllegalArgumentException("Deduplication not supported for entity type: " + dedupConf.getEntityType());
|
| 131 |
|
|
}
|
| 132 |
|
|
return oafRel;
|
| 133 |
|
|
}
|
| 134 |
|
|
|
| 135 |
|
|
private static Dedup.Builder dedup(final Dedup.RelName relClass, final String scheme) {
|
| 136 |
|
|
return Dedup.newBuilder().setRelMetadata(
|
| 137 |
|
|
RelMetadata.newBuilder().setSemantics(
|
| 138 |
|
|
Qualifier.newBuilder().setClassid(relClass.toString()).setClassname(relClass.toString()).setSchemeid(scheme).setSchemename(scheme)));
|
| 139 |
|
|
}
|
| 140 |
|
|
|
| 141 |
27941
|
claudio.at
|
class ColumnFamily {
|
| 142 |
|
|
|
| 143 |
|
|
private final RelType relType;
|
| 144 |
|
|
private final SubRelType subRelType;
|
| 145 |
|
|
|
| 146 |
28308
|
claudio.at
|
public ColumnFamily(final RelType relType, final SubRelType subRelType) {
|
| 147 |
27941
|
claudio.at
|
this.relType = relType;
|
| 148 |
|
|
this.subRelType = subRelType;
|
| 149 |
|
|
}
|
| 150 |
|
|
|
| 151 |
|
|
@Override
|
| 152 |
|
|
public String toString() {
|
| 153 |
|
|
return getRelType() + CF_SEPARATOR + getSubRelType();
|
| 154 |
|
|
}
|
| 155 |
|
|
|
| 156 |
|
|
public RelType getRelType() {
|
| 157 |
|
|
return relType;
|
| 158 |
|
|
}
|
| 159 |
|
|
|
| 160 |
|
|
public SubRelType getSubRelType() {
|
| 161 |
|
|
return subRelType;
|
| 162 |
|
|
}
|
| 163 |
|
|
|
| 164 |
|
|
}
|
| 165 |
|
|
|
| 166 |
26600
|
sandro.lab
|
}
|