1
|
package eu.dnetlib.data.mapreduce.hbase.dedup;
|
2
|
|
3
|
import java.io.IOException;
|
4
|
import java.nio.ByteBuffer;
|
5
|
import java.util.Map;
|
6
|
|
7
|
import com.google.protobuf.InvalidProtocolBufferException;
|
8
|
import eu.dnetlib.data.mapreduce.JobParams;
|
9
|
import eu.dnetlib.data.mapreduce.util.DedupUtils;
|
10
|
import eu.dnetlib.data.proto.DedupProtos.Dedup;
|
11
|
import eu.dnetlib.data.proto.KindProtos.Kind;
|
12
|
import eu.dnetlib.data.proto.OafProtos.Oaf;
|
13
|
import eu.dnetlib.data.proto.OafProtos.OafRel.Builder;
|
14
|
import eu.dnetlib.data.proto.TypeProtos.Type;
|
15
|
import eu.dnetlib.data.transform.xml.AbstractDNetXsltFunctions;
|
16
|
import eu.dnetlib.pace.config.DedupConfig;
|
17
|
import org.apache.hadoop.hbase.client.Durability;
|
18
|
import org.apache.hadoop.hbase.client.Put;
|
19
|
import org.apache.hadoop.hbase.client.Result;
|
20
|
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
|
21
|
import org.apache.hadoop.hbase.mapreduce.TableMapper;
|
22
|
import org.apache.hadoop.hbase.util.Bytes;
|
23
|
|
24
|
public class DedupFindRootsMapper extends TableMapper<ImmutableBytesWritable, Put> {
|
25
|
|
26
|
private DedupConfig dedupConf;
|
27
|
|
28
|
@Override
|
29
|
protected void setup(final Context context) throws IOException, InterruptedException {
|
30
|
dedupConf = DedupConfig.load(context.getConfiguration().get(JobParams.DEDUP_CONF));
|
31
|
System.out.println("dedup findRoots mapper\nwf conf: " + dedupConf.toString());
|
32
|
}
|
33
|
|
34
|
@Override
|
35
|
protected void map(final ImmutableBytesWritable rowkey, final Result value, final Context context) throws IOException, InterruptedException {
|
36
|
// System.out.println("Find root mapping: " + new String(rowkey.copyBytes()));
|
37
|
|
38
|
final Type type = Type.valueOf(dedupConf.getWf().getEntityType());
|
39
|
final Map<byte[], byte[]> similarRels = value.getFamilyMap(DedupUtils.getSimilarityCFBytes(type));
|
40
|
|
41
|
if ((similarRels != null) && !similarRels.isEmpty()) {
|
42
|
final ByteBuffer min = findMin(ByteBuffer.wrap(rowkey.get()), similarRels.keySet());
|
43
|
|
44
|
final byte[] row = rowkey.copyBytes();
|
45
|
final byte[] root = DedupUtils.newIdBytes(min, dedupConf.getWf().getDedupRun());
|
46
|
|
47
|
// System.out.println("Found root: " + new String(root));
|
48
|
|
49
|
emitDedupRel(context, DedupUtils.getDedupCF_mergedInBytes(type), row, root, buildRel(row, root, Dedup.RelName.isMergedIn));
|
50
|
emitDedupRel(context, DedupUtils.getDedupCF_mergesBytes(type), root, row, buildRel(root, row, Dedup.RelName.merges));
|
51
|
|
52
|
context.getCounter(dedupConf.getWf().getEntityType(), "dedupRel (x2)").increment(1);
|
53
|
|
54
|
// marks the original body deleted
|
55
|
emitBody(context, row, value.getValue(Bytes.toBytes(dedupConf.getWf().getEntityType()), DedupUtils.BODY_B));
|
56
|
|
57
|
} else {
|
58
|
context.getCounter(dedupConf.getWf().getEntityType(), "row not in similarity mesh").increment(1);
|
59
|
}
|
60
|
}
|
61
|
|
62
|
private ByteBuffer findMin(ByteBuffer min, final Iterable<byte[]> keys) {
|
63
|
for (final byte[] q : keys) {
|
64
|
final ByteBuffer iq = ByteBuffer.wrap(q);
|
65
|
if (min.compareTo(iq) > 0) {
|
66
|
min = iq;
|
67
|
}
|
68
|
}
|
69
|
return min;
|
70
|
}
|
71
|
|
72
|
private void emitBody(final Context context, final byte[] row, final byte[] body) throws InvalidProtocolBufferException, IOException, InterruptedException {
|
73
|
if (body == null) {
|
74
|
context.getCounter(dedupConf.getWf().getEntityType(), "missing body").increment(1);
|
75
|
System.err.println("missing body: " + new String(row));
|
76
|
return;
|
77
|
}
|
78
|
final Oaf prototype = Oaf.parseFrom(body);
|
79
|
|
80
|
if (prototype.getDataInfo().getDeletedbyinference()) {
|
81
|
context.getCounter(dedupConf.getWf().getEntityType(), "bodies already deleted").increment(1);
|
82
|
} else {
|
83
|
final Oaf.Builder oafRoot = Oaf.newBuilder(prototype);
|
84
|
oafRoot.getDataInfoBuilder().setDeletedbyinference(true).setInferred(true).setInferenceprovenance(dedupConf.getWf().getConfigurationId());
|
85
|
final byte[] family = Bytes.toBytes(dedupConf.getWf().getEntityType());
|
86
|
final Put put = new Put(row).add(family, DedupUtils.BODY_B, oafRoot.build().toByteArray());
|
87
|
put.setDurability(Durability.USE_DEFAULT);
|
88
|
context.write(new ImmutableBytesWritable(row), put);
|
89
|
context.getCounter(dedupConf.getWf().getEntityType(), "bodies marked deleted").increment(1);
|
90
|
}
|
91
|
}
|
92
|
|
93
|
private byte[] buildRel(final byte[] from, final byte[] to, final Dedup.RelName relClass) {
|
94
|
final Builder oafRel = DedupUtils.getDedup(dedupConf, new String(from), new String(to), relClass);
|
95
|
final Oaf oaf =
|
96
|
Oaf.newBuilder()
|
97
|
.setKind(Kind.relation)
|
98
|
.setLastupdatetimestamp(System.currentTimeMillis())
|
99
|
.setDataInfo(
|
100
|
AbstractDNetXsltFunctions.getDataInfo(null, "", "0.8", false, true).setInferenceprovenance(
|
101
|
dedupConf.getWf().getConfigurationId())).setRel(oafRel)
|
102
|
.build();
|
103
|
return oaf.toByteArray();
|
104
|
}
|
105
|
|
106
|
private void emitDedupRel(final Context context, final byte[] cf, final byte[] from, final byte[] to, final byte[] value) throws IOException,
|
107
|
InterruptedException {
|
108
|
final Put put = new Put(from).add(cf, to, value);
|
109
|
put.setWriteToWAL(JobParams.WRITE_TO_WAL);
|
110
|
context.write(new ImmutableBytesWritable(from), put);
|
111
|
}
|
112
|
|
113
|
}
|