2020-04-27 14:52:31 +02:00
|
|
|
|
2020-03-27 10:42:17 +01:00
|
|
|
package eu.dnetlib.dhp.oa.dedup;
|
2019-12-06 13:38:00 +01:00
|
|
|
|
2020-04-18 12:42:58 +02:00
|
|
|
import java.io.IOException;
|
2020-07-13 16:07:07 +02:00
|
|
|
import java.util.Optional;
|
2020-04-28 11:23:29 +02:00
|
|
|
|
2019-12-06 13:38:00 +01:00
|
|
|
import org.apache.commons.io.IOUtils;
|
2020-04-18 12:06:23 +02:00
|
|
|
import org.apache.spark.SparkConf;
|
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;
|
2020-04-17 13:12:44 +02:00
|
|
|
import org.apache.spark.sql.SaveMode;
|
2019-12-06 13:38:00 +01:00
|
|
|
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;
|
2020-04-28 11:23:29 +02:00
|
|
|
|
|
|
|
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
|
|
|
|
import eu.dnetlib.dhp.oa.dedup.model.Block;
|
|
|
|
import eu.dnetlib.dhp.schema.oaf.DataInfo;
|
|
|
|
import eu.dnetlib.dhp.schema.oaf.Relation;
|
|
|
|
import eu.dnetlib.dhp.utils.ISLookupClientFactory;
|
|
|
|
import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpException;
|
|
|
|
import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpService;
|
|
|
|
import eu.dnetlib.pace.config.DedupConfig;
|
|
|
|
import eu.dnetlib.pace.model.FieldListImpl;
|
|
|
|
import eu.dnetlib.pace.model.FieldValueImpl;
|
|
|
|
import eu.dnetlib.pace.model.MapDocument;
|
|
|
|
import eu.dnetlib.pace.util.MapDocumentUtil;
|
2019-12-06 13:38:00 +01:00
|
|
|
import scala.Tuple2;
|
|
|
|
|
2020-04-02 18:37:53 +02:00
|
|
|
public class SparkCreateSimRels extends AbstractSparkAction {
|
2019-12-06 13:38:00 +01:00
|
|
|
|
2020-04-27 14:52:31 +02:00
|
|
|
private static final Logger log = LoggerFactory.getLogger(SparkCreateSimRels.class);
|
|
|
|
|
|
|
|
public SparkCreateSimRels(ArgumentApplicationParser parser, SparkSession spark) {
|
|
|
|
super(parser, spark);
|
|
|
|
}
|
|
|
|
|
|
|
|
public static void main(String[] args) throws Exception {
|
|
|
|
ArgumentApplicationParser parser = new ArgumentApplicationParser(
|
|
|
|
IOUtils
|
|
|
|
.toString(
|
|
|
|
SparkCreateSimRels.class
|
|
|
|
.getResourceAsStream(
|
|
|
|
"/eu/dnetlib/dhp/oa/dedup/createSimRels_parameters.json")));
|
|
|
|
parser.parseArgument(args);
|
|
|
|
|
|
|
|
SparkConf conf = new SparkConf();
|
|
|
|
new SparkCreateSimRels(parser, getSparkSession(conf))
|
|
|
|
.run(ISLookupClientFactory.getLookUpService(parser.get("isLookUpUrl")));
|
|
|
|
}
|
|
|
|
|
|
|
|
@Override
|
|
|
|
public void run(ISLookUpService isLookUpService)
|
|
|
|
throws DocumentException, IOException, ISLookUpException {
|
|
|
|
|
|
|
|
// 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");
|
2020-07-13 16:07:07 +02:00
|
|
|
final int numPartitions = Optional
|
2020-07-13 16:46:13 +02:00
|
|
|
.ofNullable(parser.get("numPartitions"))
|
|
|
|
.map(Integer::valueOf)
|
|
|
|
.orElse(NUM_PARTITIONS);
|
2020-04-27 14:52:31 +02:00
|
|
|
|
2020-07-13 16:07:07 +02:00
|
|
|
log.info("numPartitions: '{}'", numPartitions);
|
2020-04-27 14:52:31 +02:00
|
|
|
log.info("graphBasePath: '{}'", graphBasePath);
|
|
|
|
log.info("isLookUpUrl: '{}'", isLookUpUrl);
|
|
|
|
log.info("actionSetId: '{}'", actionSetId);
|
|
|
|
log.info("workingPath: '{}'", workingPath);
|
|
|
|
|
|
|
|
// for each dedup configuration
|
|
|
|
for (DedupConfig dedupConf : getConfigurations(isLookUpService, actionSetId)) {
|
|
|
|
|
|
|
|
final String entity = dedupConf.getWf().getEntityType();
|
|
|
|
final String subEntity = dedupConf.getWf().getSubEntityValue();
|
|
|
|
log.info("Creating simrels for: '{}'", subEntity);
|
|
|
|
|
|
|
|
final String outputPath = DedupUtility.createSimRelPath(workingPath, actionSetId, subEntity);
|
|
|
|
removeOutputDir(spark, outputPath);
|
|
|
|
|
|
|
|
JavaSparkContext sc = JavaSparkContext.fromSparkContext(spark.sparkContext());
|
|
|
|
|
|
|
|
JavaPairRDD<String, MapDocument> mapDocuments = sc
|
|
|
|
.textFile(DedupUtility.createEntityPath(graphBasePath, subEntity))
|
2020-07-13 16:07:07 +02:00
|
|
|
.repartition(numPartitions)
|
2020-04-27 14:52:31 +02:00
|
|
|
.mapToPair(
|
|
|
|
(PairFunction<String, String, MapDocument>) s -> {
|
|
|
|
MapDocument d = MapDocumentUtil.asMapDocumentWithJPath(dedupConf, s);
|
|
|
|
return new Tuple2<>(d.getIdentifier(), d);
|
|
|
|
});
|
|
|
|
|
|
|
|
// create blocks for deduplication
|
2020-07-08 22:31:46 +02:00
|
|
|
JavaPairRDD<String, Block> blocks = Deduper
|
|
|
|
.createSortedBlocks(mapDocuments, dedupConf)
|
2020-07-13 16:07:07 +02:00
|
|
|
.repartition(numPartitions);
|
2020-04-27 14:52:31 +02:00
|
|
|
|
|
|
|
// create relations by comparing only elements in the same group
|
2020-07-08 22:31:46 +02:00
|
|
|
Deduper
|
2020-04-27 14:52:31 +02:00
|
|
|
.computeRelations(sc, blocks, dedupConf)
|
2020-07-08 22:31:46 +02:00
|
|
|
.map(t -> createSimRel(t._1(), t._2(), entity))
|
2020-07-13 16:07:07 +02:00
|
|
|
.repartition(numPartitions)
|
2020-07-08 22:31:46 +02:00
|
|
|
.map(r -> OBJECT_MAPPER.writeValueAsString(r))
|
|
|
|
.saveAsTextFile(outputPath);
|
2020-04-27 14:52:31 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
private Relation createSimRel(String source, String target, String entity) {
|
|
|
|
final Relation r = new Relation();
|
|
|
|
r.setSource(source);
|
|
|
|
r.setTarget(target);
|
|
|
|
r.setSubRelType("dedupSimilarity");
|
|
|
|
r.setRelClass("isSimilarTo");
|
|
|
|
r.setDataInfo(new DataInfo());
|
|
|
|
|
|
|
|
switch (entity) {
|
|
|
|
case "result":
|
|
|
|
r.setRelType("resultResult");
|
|
|
|
break;
|
|
|
|
case "organization":
|
|
|
|
r.setRelType("organizationOrganization");
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
throw new IllegalArgumentException("unmanaged entity type: " + entity);
|
|
|
|
}
|
|
|
|
return r;
|
|
|
|
}
|
2020-03-20 13:01:56 +01:00
|
|
|
}
|