1 |
45419
|
claudio.at
|
package eu.dnetlib.data.mapreduce.hbase.dedup;
|
2 |
|
|
|
3 |
|
|
import java.io.IOException;
|
4 |
|
|
import java.util.Collection;
|
5 |
|
|
import java.util.List;
|
6 |
|
|
import java.util.Map;
|
7 |
|
|
|
8 |
|
|
import com.google.common.collect.Maps;
|
9 |
|
|
import eu.dnetlib.data.graph.model.DNGFDecoder;
|
10 |
|
|
import eu.dnetlib.data.mapreduce.JobParams;
|
11 |
|
|
import eu.dnetlib.data.mapreduce.util.dao.HBaseTableDAO;
|
12 |
|
|
import eu.dnetlib.data.proto.DNGFProtos.DNGFEntity;
|
13 |
45791
|
claudio.at
|
import eu.dnetlib.data.proto.DliFieldTypeProtos;
|
14 |
|
|
import eu.dnetlib.data.proto.DliProtos;
|
15 |
45419
|
claudio.at
|
import eu.dnetlib.data.proto.TypeProtos.Type;
|
16 |
|
|
import eu.dnetlib.pace.clustering.BlacklistAwareClusteringCombiner;
|
17 |
|
|
import eu.dnetlib.pace.config.DedupConfig;
|
18 |
|
|
import eu.dnetlib.pace.model.MapDocument;
|
19 |
|
|
import eu.dnetlib.pace.model.ProtoDocumentBuilder;
|
20 |
|
|
import org.apache.commons.logging.Log;
|
21 |
|
|
import org.apache.commons.logging.LogFactory;
|
22 |
|
|
import org.apache.hadoop.hbase.client.Result;
|
23 |
|
|
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
|
24 |
|
|
import org.apache.hadoop.hbase.mapreduce.TableMapper;
|
25 |
|
|
import org.apache.hadoop.hbase.util.Bytes;
|
26 |
|
|
import org.apache.hadoop.io.Text;
|
27 |
|
|
|
28 |
|
|
public class DedupMapper extends TableMapper<Text, ImmutableBytesWritable> {
|
29 |
|
|
|
30 |
|
|
private static final Log log = LogFactory.getLog(DedupMapper.class);
|
31 |
|
|
|
32 |
|
|
private DedupConfig dedupConf;
|
33 |
|
|
|
34 |
|
|
private Map<String, List<String>> blackListMap = Maps.newHashMap();
|
35 |
|
|
|
36 |
|
|
private Text outKey;
|
37 |
|
|
|
38 |
|
|
private ImmutableBytesWritable ibw;
|
39 |
|
|
|
40 |
|
|
@Override
|
41 |
|
|
protected void setup(final Context context) throws IOException, InterruptedException {
|
42 |
|
|
|
43 |
|
|
final String dedupConfJson = context.getConfiguration().get(JobParams.DEDUP_CONF);
|
44 |
|
|
|
45 |
|
|
log.info("pace conf strings");
|
46 |
|
|
log.info("pace conf: " + dedupConfJson);
|
47 |
|
|
|
48 |
|
|
dedupConf = DedupConfig.load(dedupConfJson);
|
49 |
|
|
|
50 |
|
|
blackListMap = dedupConf.getPace().getBlacklists();
|
51 |
|
|
|
52 |
|
|
outKey = new Text();
|
53 |
|
|
ibw = new ImmutableBytesWritable();
|
54 |
|
|
|
55 |
|
|
log.info("pace conf");
|
56 |
|
|
log.info("entity type: " + dedupConf.getWf().getEntityType());
|
57 |
|
|
log.info("clustering: " + dedupConf.getPace().getClustering());
|
58 |
|
|
log.info("conditions: " + dedupConf.getPace().getConditions());
|
59 |
|
|
log.info("fields: " + dedupConf.getPace().getModel());
|
60 |
|
|
log.info("blacklists: " + blackListMap);
|
61 |
|
|
log.info("wf conf: " + dedupConf.toString());
|
62 |
|
|
}
|
63 |
|
|
|
64 |
|
|
@Override
|
65 |
|
|
protected void map(final ImmutableBytesWritable keyIn, final Result result, final Context context) throws IOException, InterruptedException {
|
66 |
|
|
// log.info("got key: " + new String(keyIn.copyBytes()));
|
67 |
|
|
|
68 |
45705
|
claudio.at
|
final byte[] body = result.getValue(HBaseTableDAO.cfMetadataByte(), dedupConf.getWf().getEntityType().getBytes());
|
69 |
45419
|
claudio.at
|
|
70 |
|
|
if (body != null) {
|
71 |
|
|
|
72 |
45791
|
claudio.at
|
final DNGFDecoder decoder = DNGFDecoder.decode(body, DliFieldTypeProtos.completionStatus, DliProtos.completionStatus, DliProtos.resolvedfrom, DliProtos.typedIdentifier);
|
73 |
45419
|
claudio.at
|
if (decoder.getDNGF().getDataInfo().getDeletedbyinference()) {
|
74 |
|
|
context.getCounter(dedupConf.getWf().getEntityType(), "deleted by inference").increment(1);
|
75 |
|
|
return;
|
76 |
|
|
}
|
77 |
|
|
|
78 |
|
|
final DNGFEntity entity = decoder.getEntity();
|
79 |
|
|
|
80 |
|
|
context.getCounter(entity.getType().toString(), "decoded").increment(1);
|
81 |
|
|
|
82 |
|
|
if (entity.getType().equals(Type.valueOf(dedupConf.getWf().getEntityType()))) {
|
83 |
|
|
|
84 |
|
|
// GeneratedMessage metadata = DNGFEntityDecoder.decode(entity).getEntity();
|
85 |
|
|
final MapDocument doc = ProtoDocumentBuilder.newInstance(Bytes.toString(keyIn.copyBytes()), entity, dedupConf.getPace().getModel());
|
86 |
|
|
emitNGrams(context, doc, BlacklistAwareClusteringCombiner.filterAndCombine(doc, dedupConf, blackListMap));
|
87 |
|
|
}
|
88 |
|
|
} else {
|
89 |
|
|
context.getCounter(dedupConf.getWf().getEntityType(), "missing body").increment(1);
|
90 |
|
|
}
|
91 |
|
|
}
|
92 |
|
|
|
93 |
|
|
private void emitNGrams(final Context context, final MapDocument doc, final Collection<String> ngrams) throws IOException, InterruptedException {
|
94 |
|
|
for (final String ngram : ngrams) {
|
95 |
|
|
outKey.set(ngram);
|
96 |
|
|
ibw.set(doc.toByteArray());
|
97 |
|
|
context.write(outKey, ibw);
|
98 |
|
|
}
|
99 |
|
|
}
|
100 |
|
|
|
101 |
|
|
}
|