2020-03-27 10:42:17 +01:00
|
|
|
package eu.dnetlib.dhp.oa.dedup;
|
2019-12-06 13:38:00 +01:00
|
|
|
|
2020-03-20 13:01:56 +01:00
|
|
|
import com.fasterxml.jackson.core.JsonProcessingException;
|
|
|
|
import com.fasterxml.jackson.databind.ObjectMapper;
|
2019-12-06 13:38:00 +01:00
|
|
|
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
|
2020-03-20 13:01:56 +01:00
|
|
|
import eu.dnetlib.dhp.schema.action.AtomicAction;
|
2019-12-06 13:38:00 +01:00
|
|
|
import eu.dnetlib.dhp.schema.oaf.Relation;
|
2020-04-01 18:48:31 +02:00
|
|
|
import eu.dnetlib.dhp.utils.ISLookupClientFactory;
|
2020-03-20 13:01:56 +01:00
|
|
|
import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpException;
|
2020-04-01 18:48:31 +02:00
|
|
|
import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpService;
|
2019-12-06 13:38:00 +01:00
|
|
|
import eu.dnetlib.pace.config.DedupConfig;
|
|
|
|
import eu.dnetlib.pace.model.MapDocument;
|
|
|
|
import eu.dnetlib.pace.util.MapDocumentUtil;
|
|
|
|
import org.apache.commons.io.IOUtils;
|
2020-03-20 13:01:56 +01:00
|
|
|
import org.apache.hadoop.io.Text;
|
2019-12-06 13:38:00 +01:00
|
|
|
import org.apache.spark.api.java.JavaPairRDD;
|
|
|
|
import org.apache.spark.api.java.JavaRDD;
|
|
|
|
import org.apache.spark.api.java.JavaSparkContext;
|
2020-04-01 18:48:31 +02:00
|
|
|
import org.apache.spark.api.java.function.PairFunction;
|
2019-12-06 13:38:00 +01:00
|
|
|
import org.apache.spark.sql.Encoders;
|
|
|
|
import org.apache.spark.sql.SparkSession;
|
2020-03-20 13:01:56 +01:00
|
|
|
import org.dom4j.DocumentException;
|
2020-04-16 12:36:37 +02:00
|
|
|
import org.slf4j.Logger;
|
|
|
|
import org.slf4j.LoggerFactory;
|
2019-12-06 13:38:00 +01:00
|
|
|
import scala.Tuple2;
|
|
|
|
|
2020-04-02 18:37:53 +02:00
|
|
|
import java.io.IOException;
|
2019-12-06 13:38:00 +01:00
|
|
|
import java.util.List;
|
|
|
|
|
2020-04-02 18:37:53 +02:00
|
|
|
public class SparkCreateSimRels extends AbstractSparkAction {
|
2019-12-06 13:38:00 +01:00
|
|
|
|
2020-04-16 12:36:37 +02:00
|
|
|
private static final Logger log = LoggerFactory.getLogger(SparkCreateSimRels.class);
|
2019-12-06 13:38:00 +01:00
|
|
|
|
2020-04-16 12:36:37 +02:00
|
|
|
public SparkCreateSimRels(ArgumentApplicationParser parser, SparkSession spark) {
|
2020-04-03 18:32:25 +02:00
|
|
|
super(parser, spark);
|
2020-04-02 18:37:53 +02:00
|
|
|
}
|
|
|
|
|
2019-12-06 13:38:00 +01:00
|
|
|
public static void main(String[] args) throws Exception {
|
2020-04-02 18:37:53 +02:00
|
|
|
ArgumentApplicationParser parser = new ArgumentApplicationParser(
|
2020-03-27 10:42:17 +01:00
|
|
|
IOUtils.toString(
|
|
|
|
SparkCreateSimRels.class.getResourceAsStream("/eu/dnetlib/dhp/oa/dedup/createSimRels_parameters.json")));
|
2019-12-06 13:38:00 +01:00
|
|
|
parser.parseArgument(args);
|
|
|
|
|
2020-04-03 18:32:25 +02:00
|
|
|
new SparkCreateSimRels(parser, getSparkSession(parser)).run(ISLookupClientFactory.getLookUpService(parser.get("isLookUpUrl")));
|
2020-03-20 13:01:56 +01:00
|
|
|
}
|
|
|
|
|
2020-04-02 18:37:53 +02:00
|
|
|
@Override
|
2020-04-03 18:32:25 +02:00
|
|
|
public void run(ISLookUpService isLookUpService) throws DocumentException, IOException, ISLookUpException {
|
2020-03-20 13:01:56 +01:00
|
|
|
|
|
|
|
//read oozie parameters
|
|
|
|
final String graphBasePath = parser.get("graphBasePath");
|
|
|
|
final String isLookUpUrl = parser.get("isLookUpUrl");
|
|
|
|
final String actionSetId = parser.get("actionSetId");
|
|
|
|
final String workingPath = parser.get("workingPath");
|
|
|
|
|
2020-03-20 19:17:24 +01:00
|
|
|
System.out.println(String.format("graphBasePath: '%s'", graphBasePath));
|
2020-04-15 16:34:06 +02:00
|
|
|
System.out.println(String.format("isLookUpUrl: '%s'", isLookUpUrl));
|
|
|
|
System.out.println(String.format("actionSetId: '%s'", actionSetId));
|
|
|
|
System.out.println(String.format("workingPath: '%s'", workingPath));
|
2020-03-20 19:17:24 +01:00
|
|
|
|
2020-04-03 18:32:25 +02:00
|
|
|
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
|
2020-04-02 18:37:53 +02:00
|
|
|
|
2020-04-03 18:32:25 +02:00
|
|
|
//for each dedup configuration
|
|
|
|
for (DedupConfig dedupConf: getConfigurations(isLookUpService, actionSetId)) {
|
2020-04-15 16:34:06 +02:00
|
|
|
|
2020-04-03 18:32:25 +02:00
|
|
|
final String entity = dedupConf.getWf().getEntityType();
|
|
|
|
final String subEntity = dedupConf.getWf().getSubEntityValue();
|
2020-04-15 16:34:06 +02:00
|
|
|
System.out.println(String.format("Creating simrels for: '%s'", subEntity));
|
2020-04-02 18:37:53 +02:00
|
|
|
|
2020-04-03 18:32:25 +02:00
|
|
|
JavaPairRDD<String, MapDocument> mapDocument = sc.textFile(DedupUtility.createEntityPath(graphBasePath, subEntity))
|
|
|
|
.mapToPair((PairFunction<String, String, MapDocument>) s -> {
|
|
|
|
MapDocument d = MapDocumentUtil.asMapDocumentWithJPath(dedupConf, s);
|
|
|
|
return new Tuple2<String, MapDocument>(d.getIdentifier(), d);
|
|
|
|
});
|
2020-04-02 18:37:53 +02:00
|
|
|
|
2020-04-03 18:32:25 +02:00
|
|
|
//create blocks for deduplication
|
2020-04-16 12:36:37 +02:00
|
|
|
JavaPairRDD<String, List<MapDocument>> blocks = Deduper.createSortedBlocks(sc, mapDocument, dedupConf);
|
2020-04-02 18:37:53 +02:00
|
|
|
|
2020-04-03 18:32:25 +02:00
|
|
|
//create relations by comparing only elements in the same group
|
2020-04-16 12:36:37 +02:00
|
|
|
final JavaPairRDD<String, String> dedupRels = Deduper.computeRelations(sc, blocks, dedupConf);
|
2020-04-02 18:37:53 +02:00
|
|
|
|
2020-04-03 18:32:25 +02:00
|
|
|
JavaRDD<Relation> relationsRDD = dedupRels.map(r -> createSimRel(r._1(), r._2(), entity));
|
2020-04-02 18:37:53 +02:00
|
|
|
|
2020-04-03 18:32:25 +02:00
|
|
|
//save the simrel in the workingdir
|
|
|
|
spark.createDataset(relationsRDD.rdd(), Encoders.bean(Relation.class))
|
|
|
|
.write()
|
|
|
|
.mode("overwrite")
|
|
|
|
.save(DedupUtility.createSimRelPath(workingPath, actionSetId, subEntity));
|
2020-04-02 18:37:53 +02:00
|
|
|
}
|
2020-03-20 13:01:56 +01:00
|
|
|
}
|
|
|
|
|
2020-03-25 14:16:06 +01:00
|
|
|
/**
|
|
|
|
* Utility method used to create an atomic action from a Relation object
|
|
|
|
* @param relation input relation
|
|
|
|
* @return A tuple2 with [id, json serialization of the atomic action]
|
|
|
|
* @throws JsonProcessingException
|
|
|
|
*/
|
2020-03-20 13:01:56 +01:00
|
|
|
public Tuple2<Text, Text> createSequenceFileRow(Relation relation) throws JsonProcessingException {
|
|
|
|
|
|
|
|
ObjectMapper mapper = new ObjectMapper();
|
|
|
|
|
|
|
|
String id = relation.getSource() + "@" + relation.getRelClass() + "@" + relation.getTarget();
|
|
|
|
AtomicAction<Relation> aa = new AtomicAction<>(Relation.class, relation);
|
|
|
|
|
|
|
|
return new Tuple2<>(
|
|
|
|
new Text(id),
|
|
|
|
new Text(mapper.writeValueAsString(aa))
|
|
|
|
);
|
|
|
|
}
|
|
|
|
|
|
|
|
public Relation createSimRel(String source, String target, String entity){
|
|
|
|
final Relation r = new Relation();
|
|
|
|
r.setSource(source);
|
|
|
|
r.setTarget(target);
|
2019-12-06 13:38:00 +01:00
|
|
|
|
2020-03-20 13:01:56 +01:00
|
|
|
switch(entity){
|
|
|
|
case "result":
|
|
|
|
r.setRelClass("resultResult_dedupSimilarity_isSimilarTo");
|
|
|
|
break;
|
|
|
|
case "organization":
|
|
|
|
r.setRelClass("organizationOrganization_dedupSimilarity_isSimilarTo");
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
r.setRelClass("isSimilarTo");
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return r;
|
|
|
|
}
|
|
|
|
}
|