|
1
|
package eu.dnetlib.data.mapreduce.util;
|
|
2
|
|
|
3
|
import java.nio.ByteBuffer;
|
|
4
|
|
|
5
|
import eu.dnetlib.data.mapreduce.Algorithms;
|
|
6
|
import eu.dnetlib.data.mapreduce.JobParams;
|
|
7
|
import eu.dnetlib.data.proto.DedupProtos.Dedup;
|
|
8
|
import eu.dnetlib.data.proto.DedupSimilarityProtos.DedupSimilarity;
|
|
9
|
import eu.dnetlib.data.proto.FieldTypeProtos.Qualifier;
|
|
10
|
import eu.dnetlib.data.proto.KindProtos.Kind;
|
|
11
|
import eu.dnetlib.data.proto.OafProtos.Oaf;
|
|
12
|
import eu.dnetlib.data.proto.OafProtos.OafRel;
|
|
13
|
import eu.dnetlib.data.proto.OafProtos.OafRel.Builder;
|
|
14
|
import eu.dnetlib.data.proto.OrganizationOrganizationProtos.OrganizationOrganization;
|
|
15
|
import eu.dnetlib.data.proto.RelMetadataProtos.RelMetadata;
|
|
16
|
import eu.dnetlib.data.proto.RelTypeProtos.RelType;
|
|
17
|
import eu.dnetlib.data.proto.RelTypeProtos.SubRelType;
|
|
18
|
import eu.dnetlib.data.proto.ResultResultProtos.ResultResult;
|
|
19
|
import eu.dnetlib.data.proto.TypeProtos.Type;
|
|
20
|
import eu.dnetlib.data.transform.xml.AbstractDNetXsltFunctions;
|
|
21
|
import eu.dnetlib.pace.config.DedupConfig;
|
|
22
|
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
|
|
23
|
import org.apache.hadoop.hbase.util.Bytes;
|
|
24
|
|
|
25
|
public class DedupUtils {
|
|
26
|
|
|
27
|
public static final String CF_SEPARATOR = "_";
|
|
28
|
|
|
29
|
public static final String ROOT = "dedup_wf";
|
|
30
|
|
|
31
|
public static final String BODY_S = "body";
|
|
32
|
|
|
33
|
public static final byte[] BODY_B = Bytes.toBytes(BODY_S);
|
|
34
|
|
|
35
|
public static String dedupPrefix(final String dedupRun) {
|
|
36
|
return "|" + ROOT + "_" + dedupRun + "::";
|
|
37
|
}
|
|
38
|
|
|
39
|
public static String newId(final String id, final String dedupRun) {
|
|
40
|
if ((dedupRun == null) || (dedupRun.length() != 3)) throw new IllegalArgumentException("wrong dedupRun param");
|
|
41
|
|
|
42
|
return id.replaceFirst("\\|.*\\:\\:", dedupPrefix(dedupRun));
|
|
43
|
}
|
|
44
|
|
|
45
|
public static byte[] newIdBytes(final String s, final String dedupRun) {
|
|
46
|
return newId(s, dedupRun).getBytes();
|
|
47
|
}
|
|
48
|
|
|
49
|
public static byte[] newIdBytes(final ByteBuffer b, final String dedupRun) {
|
|
50
|
return newId(new String(b.array()), dedupRun).getBytes();
|
|
51
|
}
|
|
52
|
|
|
53
|
public static boolean isRoot(final String s) {
|
|
54
|
return s.contains(ROOT);
|
|
55
|
}
|
|
56
|
|
|
57
|
public static boolean isRoot(final ImmutableBytesWritable s) {
|
|
58
|
return isRoot(s.copyBytes());
|
|
59
|
}
|
|
60
|
|
|
61
|
public static boolean isRoot(final byte[] s) {
|
|
62
|
return isRoot(new String(s));
|
|
63
|
}
|
|
64
|
|
|
65
|
public static String getDedupCF_merges(final Type type) {
|
|
66
|
return getRelType(type) + CF_SEPARATOR + SubRelType.dedup + CF_SEPARATOR + Dedup.RelName.merges;
|
|
67
|
}
|
|
68
|
|
|
69
|
public static String getDedupCF_merges(final String type) {
|
|
70
|
return getDedupCF_merges(Type.valueOf(type));
|
|
71
|
}
|
|
72
|
|
|
73
|
public static byte[] getDedupCF_mergesBytes(final Type type) {
|
|
74
|
return Bytes.toBytes(getDedupCF_merges(type));
|
|
75
|
}
|
|
76
|
|
|
77
|
public static byte[] getDedupCF_mergesBytes(final String type) {
|
|
78
|
return getDedupCF_mergesBytes(Type.valueOf(type));
|
|
79
|
}
|
|
80
|
|
|
81
|
public static String getDedupCF_mergedIn(final Type type) {
|
|
82
|
return getRelType(type) + CF_SEPARATOR + SubRelType.dedup + CF_SEPARATOR + Dedup.RelName.isMergedIn;
|
|
83
|
}
|
|
84
|
|
|
85
|
public static String getDedupCF_mergedIn(final String type) {
|
|
86
|
return getDedupCF_mergedIn(Type.valueOf(type));
|
|
87
|
}
|
|
88
|
|
|
89
|
public static byte[] getDedupCF_mergedInBytes(final Type type) {
|
|
90
|
return Bytes.toBytes(getDedupCF_mergedIn(type));
|
|
91
|
}
|
|
92
|
|
|
93
|
public static byte[] getDedupCF_mergedInBytes(final String type) {
|
|
94
|
return getDedupCF_mergedInBytes(Type.valueOf(type));
|
|
95
|
}
|
|
96
|
|
|
97
|
public static String getSimilarityCF(final Type type) {
|
|
98
|
return getRelType(type) + CF_SEPARATOR + SubRelType.dedupSimilarity + CF_SEPARATOR + DedupSimilarity.RelName.isSimilarTo;
|
|
99
|
}
|
|
100
|
|
|
101
|
public static String getSimilarityCF(final String type) {
|
|
102
|
return getSimilarityCF(Type.valueOf(type));
|
|
103
|
}
|
|
104
|
|
|
105
|
public static byte[] getSimilarityCFBytes(final Type type) {
|
|
106
|
return Bytes.toBytes(getSimilarityCF(type));
|
|
107
|
}
|
|
108
|
|
|
109
|
public static byte[] getSimilarityCFBytes(final String type) {
|
|
110
|
return getSimilarityCFBytes(Type.valueOf(type));
|
|
111
|
}
|
|
112
|
|
|
113
|
public static String getRelTypeString(final Type type) {
|
|
114
|
return getRelType(type).toString();
|
|
115
|
}
|
|
116
|
|
|
117
|
public static RelType getRelType(final Type type) {
|
|
118
|
switch (type) {
|
|
119
|
case organization:
|
|
120
|
return RelType.organizationOrganization;
|
|
121
|
case result:
|
|
122
|
return RelType.resultResult;
|
|
123
|
default:
|
|
124
|
throw new IllegalArgumentException("Deduplication not supported for entity type: " + type);
|
|
125
|
}
|
|
126
|
}
|
|
127
|
|
|
128
|
public static ColumnFamily decodeCF(final byte[] b) {
|
|
129
|
final String[] s = new String(b).split(CF_SEPARATOR);
|
|
130
|
return new DedupUtils().getCF(RelType.valueOf(s[0]), SubRelType.valueOf(s[1]));
|
|
131
|
}
|
|
132
|
|
|
133
|
private ColumnFamily getCF(final RelType relType, final SubRelType subRelType) {
|
|
134
|
return new ColumnFamily(relType, subRelType);
|
|
135
|
}
|
|
136
|
|
|
137
|
public static OafRel.Builder getDedup(final DedupConfig dedupConf, final String from, final String to, final Dedup.RelName relClass) {
|
|
138
|
final Type type = Type.valueOf(dedupConf.getWf().getEntityType());
|
|
139
|
final Builder oafRel = getRelBuilder(from, to, relClass.name(), DedupUtils.getRelType(type));
|
|
140
|
switch (type) {
|
|
141
|
case organization:
|
|
142
|
oafRel.setOrganizationOrganization(OrganizationOrganization.newBuilder().setDedup(
|
|
143
|
dedup(relClass.name(), "dnet:organization_organization_relations")));
|
|
144
|
break;
|
|
145
|
case result:
|
|
146
|
oafRel.setResultResult(ResultResult.newBuilder().setDedup(DedupUtils.dedup(relClass.name(), "dnet:result_result_relations")));
|
|
147
|
break;
|
|
148
|
default:
|
|
149
|
throw new IllegalArgumentException("Deduplication not supported for entity type: " + dedupConf.getWf().getEntityType());
|
|
150
|
}
|
|
151
|
return oafRel;
|
|
152
|
}
|
|
153
|
|
|
154
|
public static OafRel.Builder getDedupSimilarity(final DedupConfig dedupConf, final String from, final String to) {
|
|
155
|
final Type type = Type.valueOf(dedupConf.getWf().getEntityType());
|
|
156
|
final String isSimilarTo = DedupSimilarity.RelName.isSimilarTo.name();
|
|
157
|
final Builder oafRel = getRelBuilder(from, to, isSimilarTo, DedupUtils.getRelType(type));
|
|
158
|
switch (type) {
|
|
159
|
case organization:
|
|
160
|
oafRel.setOrganizationOrganization(OrganizationOrganization.newBuilder().setDedupSimilarity(
|
|
161
|
dedupSimilarity(isSimilarTo, "dnet:organization_organization_relations")));
|
|
162
|
break;
|
|
163
|
case result:
|
|
164
|
oafRel.setResultResult(ResultResult.newBuilder().setDedupSimilarity(dedupSimilarity(isSimilarTo, "dnet:result_result_relations")));
|
|
165
|
break;
|
|
166
|
default:
|
|
167
|
throw new IllegalArgumentException("Deduplication not supported for entity type: " + dedupConf.getWf().getEntityType());
|
|
168
|
}
|
|
169
|
return oafRel;
|
|
170
|
}
|
|
171
|
|
|
172
|
private static Builder getRelBuilder(final String from, final String to, final String relClass, final RelType relType) {
|
|
173
|
return OafRel.newBuilder().setRelType(relType).setSubRelType(SubRelType.dedup).setRelClass(relClass).setChild(false)
|
|
174
|
.setSource(new String(from)).setTarget(new String(to));
|
|
175
|
}
|
|
176
|
|
|
177
|
private static Dedup.Builder dedup(final String relClass, final String scheme) {
|
|
178
|
return Dedup.newBuilder().setRelMetadata(getRelMetadata(relClass, scheme));
|
|
179
|
}
|
|
180
|
|
|
181
|
private static DedupSimilarity.Builder dedupSimilarity(final String relClass, final String scheme) {
|
|
182
|
return DedupSimilarity.newBuilder().setRelMetadata(getRelMetadata(relClass, scheme));
|
|
183
|
}
|
|
184
|
|
|
185
|
private static RelMetadata.Builder getRelMetadata(final String relClass, final String scheme) {
|
|
186
|
return RelMetadata.newBuilder().setSemantics(
|
|
187
|
Qualifier.newBuilder().setClassid(relClass).setClassname(relClass).setSchemeid(scheme).setSchemename(scheme));
|
|
188
|
}
|
|
189
|
|
|
190
|
public static Oaf.Builder buildRel(final DedupConfig dedupConf, final OafRel.Builder oafRel, final double trust) {
|
|
191
|
final double sTrust = Algorithms.scale(trust, JobParams.MAX_DEDUP_TRUST);
|
|
192
|
final Oaf.Builder oaf =
|
|
193
|
Oaf.newBuilder()
|
|
194
|
.setKind(Kind.relation)
|
|
195
|
.setLastupdatetimestamp(System.currentTimeMillis())
|
|
196
|
.setDataInfo(
|
|
197
|
AbstractDNetXsltFunctions.getDataInfo(null, "", String.valueOf(sTrust), false, true).setInferenceprovenance(
|
|
198
|
dedupConf.getWf().getConfigurationId())).setRel(oafRel);
|
|
199
|
return oaf;
|
|
200
|
}
|
|
201
|
|
|
202
|
class ColumnFamily {
|
|
203
|
|
|
204
|
private final RelType relType;
|
|
205
|
private final SubRelType subRelType;
|
|
206
|
|
|
207
|
public ColumnFamily(final RelType relType, final SubRelType subRelType) {
|
|
208
|
this.relType = relType;
|
|
209
|
this.subRelType = subRelType;
|
|
210
|
}
|
|
211
|
|
|
212
|
@Override
|
|
213
|
public String toString() {
|
|
214
|
return getRelType() + CF_SEPARATOR + getSubRelType();
|
|
215
|
}
|
|
216
|
|
|
217
|
public RelType getRelType() {
|
|
218
|
return relType;
|
|
219
|
}
|
|
220
|
|
|
221
|
public SubRelType getSubRelType() {
|
|
222
|
return subRelType;
|
|
223
|
}
|
|
224
|
|
|
225
|
}
|
|
226
|
|
|
227
|
}
|