1 |
26600
|
sandro.lab
|
package eu.dnetlib.data.mapreduce.hbase.dedup;
|
2 |
|
|
|
3 |
|
|
import java.io.IOException;
|
4 |
|
|
import java.util.Map;
|
5 |
|
|
|
6 |
36670
|
claudio.at
|
import eu.dnetlib.data.mapreduce.JobParams;
|
7 |
28094
|
claudio.at
|
import eu.dnetlib.data.mapreduce.util.DedupUtils;
|
8 |
|
|
import eu.dnetlib.data.proto.DedupProtos.Dedup;
|
9 |
26600
|
sandro.lab
|
import eu.dnetlib.data.proto.KindProtos.Kind;
|
10 |
|
|
import eu.dnetlib.data.proto.OafProtos.Oaf;
|
11 |
|
|
import eu.dnetlib.data.proto.OafProtos.OafRel.Builder;
|
12 |
28411
|
claudio.at
|
import eu.dnetlib.data.proto.TypeProtos.Type;
|
13 |
40205
|
claudio.at
|
import eu.dnetlib.data.transform.xml.AbstractDNetXsltFunctions;
|
14 |
36670
|
claudio.at
|
import eu.dnetlib.pace.config.DedupConfig;
|
15 |
39616
|
claudio.at
|
import org.apache.commons.logging.Log;
|
16 |
|
|
import org.apache.commons.logging.LogFactory;
|
17 |
57443
|
claudio.at
|
import org.apache.hadoop.hbase.client.Durability;
|
18 |
39616
|
claudio.at
|
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 |
26600
|
sandro.lab
|
|
24 |
39616
|
claudio.at
|
public class DedupMarkDeletedEntityMapper extends TableMapper<ImmutableBytesWritable, Put> {
|
25 |
26600
|
sandro.lab
|
|
26 |
39616
|
claudio.at
|
private static final Log log = LogFactory.getLog(DedupMarkDeletedEntityMapper.class);
|
27 |
|
|
|
28 |
26600
|
sandro.lab
|
private DedupConfig dedupConf;
|
29 |
|
|
|
30 |
|
|
@Override
|
31 |
28308
|
claudio.at
|
protected void setup(final Context context) throws IOException, InterruptedException {
|
32 |
36670
|
claudio.at
|
dedupConf = DedupConfig.load(context.getConfiguration().get(JobParams.DEDUP_CONF));
|
33 |
39616
|
claudio.at
|
log.info("dedup findRoots mapper\nwf conf: " + dedupConf.toString());
|
34 |
26600
|
sandro.lab
|
}
|
35 |
|
|
|
36 |
|
|
@Override
|
37 |
28308
|
claudio.at
|
protected void map(final ImmutableBytesWritable rowkey, final Result value, final Context context) throws IOException, InterruptedException {
|
38 |
26600
|
sandro.lab
|
// System.out.println("Find root mapping: " + new String(rowkey.copyBytes()));
|
39 |
|
|
|
40 |
36670
|
claudio.at
|
final Type type = Type.valueOf(dedupConf.getWf().getEntityType());
|
41 |
39616
|
claudio.at
|
final Map<byte[], byte[]> mergedIn = value.getFamilyMap(DedupUtils.getDedupCF_mergedInBytes(type));
|
42 |
26600
|
sandro.lab
|
|
43 |
39616
|
claudio.at
|
if ((mergedIn != null) && !mergedIn.isEmpty()) {
|
44 |
26600
|
sandro.lab
|
|
45 |
36670
|
claudio.at
|
final byte[] row = rowkey.copyBytes();
|
46 |
26600
|
sandro.lab
|
|
47 |
|
|
// marks the original body deleted
|
48 |
39616
|
claudio.at
|
emitBody(context, row, value.getValue(Bytes.toBytes(type.toString()), DedupUtils.BODY_B));
|
49 |
26600
|
sandro.lab
|
|
50 |
|
|
} else {
|
51 |
39616
|
claudio.at
|
context.getCounter(type.toString(), "row not merged").increment(1);
|
52 |
26600
|
sandro.lab
|
}
|
53 |
|
|
}
|
54 |
|
|
|
55 |
|
|
|
56 |
39616
|
claudio.at
|
private void emitBody(final Context context, final byte[] row, final byte[] body) throws IOException, InterruptedException {
|
57 |
|
|
final String type = dedupConf.getWf().getEntityType();
|
58 |
26600
|
sandro.lab
|
if (body == null) {
|
59 |
39616
|
claudio.at
|
context.getCounter(type, "missing body").increment(1);
|
60 |
26600
|
sandro.lab
|
System.err.println("missing body: " + new String(row));
|
61 |
|
|
return;
|
62 |
|
|
}
|
63 |
|
|
final Oaf prototype = Oaf.parseFrom(body);
|
64 |
|
|
|
65 |
|
|
if (prototype.getDataInfo().getDeletedbyinference()) {
|
66 |
39616
|
claudio.at
|
context.getCounter(type, "bodies already deleted").increment(1);
|
67 |
26600
|
sandro.lab
|
} else {
|
68 |
36670
|
claudio.at
|
final Oaf.Builder oafRoot = Oaf.newBuilder(prototype);
|
69 |
37824
|
claudio.at
|
oafRoot.getDataInfoBuilder().setDeletedbyinference(true).setInferred(true).setInferenceprovenance(dedupConf.getWf().getConfigurationId());
|
70 |
39616
|
claudio.at
|
final byte[] family = Bytes.toBytes(type);
|
71 |
36670
|
claudio.at
|
final Put put = new Put(row).add(family, DedupUtils.BODY_B, oafRoot.build().toByteArray());
|
72 |
57443
|
claudio.at
|
put.setDurability(Durability.USE_DEFAULT);
|
73 |
26600
|
sandro.lab
|
context.write(new ImmutableBytesWritable(row), put);
|
74 |
39616
|
claudio.at
|
context.getCounter(type, "bodies marked deleted").increment(1);
|
75 |
26600
|
sandro.lab
|
}
|
76 |
|
|
}
|
77 |
|
|
|
78 |
28308
|
claudio.at
|
private byte[] buildRel(final byte[] from, final byte[] to, final Dedup.RelName relClass) {
|
79 |
36670
|
claudio.at
|
final Builder oafRel = DedupUtils.getDedup(dedupConf, new String(from), new String(to), relClass);
|
80 |
|
|
final Oaf oaf =
|
81 |
37824
|
claudio.at
|
Oaf.newBuilder()
|
82 |
|
|
.setKind(Kind.relation)
|
83 |
40314
|
claudio.at
|
.setLastupdatetimestamp(System.currentTimeMillis())
|
84 |
37824
|
claudio.at
|
.setDataInfo(
|
85 |
40205
|
claudio.at
|
AbstractDNetXsltFunctions.getDataInfo(null, "", "0.8", false, true).setInferenceprovenance(
|
86 |
37824
|
claudio.at
|
dedupConf.getWf().getConfigurationId())).setRel(oafRel)
|
87 |
|
|
.build();
|
88 |
26600
|
sandro.lab
|
return oaf.toByteArray();
|
89 |
|
|
}
|
90 |
|
|
|
91 |
|
|
}
|