2020-04-27 14:52:31 +02:00
|
|
|
|
2020-03-27 10:42:17 +01:00
|
|
|
package eu.dnetlib.dhp.oa.dedup;
|
2020-03-20 13:01:56 +01:00
|
|
|
|
2020-04-18 12:42:58 +02:00
|
|
|
import static org.apache.spark.sql.functions.col;
|
|
|
|
|
2020-11-25 14:55:32 +01:00
|
|
|
import java.util.Objects;
|
|
|
|
|
2020-03-20 13:01:56 +01:00
|
|
|
import org.apache.commons.io.IOUtils;
|
2020-07-13 15:30:57 +02:00
|
|
|
import org.apache.commons.lang3.StringUtils;
|
2020-04-18 12:06:23 +02:00
|
|
|
import org.apache.spark.SparkConf;
|
2020-07-13 15:30:57 +02:00
|
|
|
import org.apache.spark.api.java.function.FilterFunction;
|
2020-04-16 11:13:51 +02:00
|
|
|
import org.apache.spark.api.java.function.MapFunction;
|
2020-04-18 12:06:23 +02:00
|
|
|
import org.apache.spark.sql.*;
|
2020-04-16 11:13:51 +02:00
|
|
|
import org.slf4j.Logger;
|
|
|
|
import org.slf4j.LoggerFactory;
|
2020-04-28 11:23:29 +02:00
|
|
|
|
|
|
|
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
|
2021-03-31 17:07:13 +02:00
|
|
|
import eu.dnetlib.dhp.schema.common.ModelConstants;
|
2020-04-28 11:23:29 +02:00
|
|
|
import eu.dnetlib.dhp.schema.common.ModelSupport;
|
|
|
|
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.ISLookUpService;
|
2020-03-20 13:01:56 +01:00
|
|
|
import scala.Tuple2;
|
2021-04-16 16:40:42 +02:00
|
|
|
import scala.Tuple3;
|
2020-03-20 13:01:56 +01:00
|
|
|
|
2020-04-06 16:30:31 +02:00
|
|
|
public class SparkPropagateRelation extends AbstractSparkAction {
|
2020-03-20 13:01:56 +01:00
|
|
|
|
2020-04-27 14:52:31 +02:00
|
|
|
private static final Logger log = LoggerFactory.getLogger(SparkPropagateRelation.class);
|
|
|
|
|
|
|
|
enum FieldType {
|
|
|
|
SOURCE, TARGET
|
|
|
|
}
|
|
|
|
|
2020-11-25 14:55:32 +01:00
|
|
|
public SparkPropagateRelation(ArgumentApplicationParser parser, SparkSession spark) {
|
2020-04-27 14:52:31 +02:00
|
|
|
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/propagateRelation_parameters.json")));
|
|
|
|
|
|
|
|
parser.parseArgument(args);
|
|
|
|
|
|
|
|
SparkConf conf = new SparkConf();
|
|
|
|
conf.set("spark.serializer", "org.apache.spark.serializer.KryoSerializer");
|
|
|
|
conf.registerKryoClasses(ModelSupport.getOafModelClasses());
|
|
|
|
|
|
|
|
new SparkPropagateRelation(parser, getSparkSession(conf))
|
|
|
|
.run(ISLookupClientFactory.getLookUpService(parser.get("isLookUpUrl")));
|
|
|
|
}
|
|
|
|
|
|
|
|
@Override
|
|
|
|
public void run(ISLookUpService isLookUpService) {
|
|
|
|
|
|
|
|
final String graphBasePath = parser.get("graphBasePath");
|
|
|
|
final String workingPath = parser.get("workingPath");
|
2020-11-25 14:55:32 +01:00
|
|
|
final String graphOutputPath = parser.get("graphOutputPath");
|
2020-04-27 14:52:31 +02:00
|
|
|
|
|
|
|
log.info("graphBasePath: '{}'", graphBasePath);
|
|
|
|
log.info("workingPath: '{}'", workingPath);
|
2020-11-25 14:55:32 +01:00
|
|
|
log.info("graphOutputPath: '{}'", graphOutputPath);
|
2020-04-27 14:52:31 +02:00
|
|
|
|
2020-11-25 14:55:32 +01:00
|
|
|
final String outputRelationPath = DedupUtility.createEntityPath(graphOutputPath, "relation");
|
2020-04-27 14:52:31 +02:00
|
|
|
removeOutputDir(spark, outputRelationPath);
|
|
|
|
|
|
|
|
Dataset<Relation> mergeRels = spark
|
|
|
|
.read()
|
|
|
|
.load(DedupUtility.createMergeRelPath(workingPath, "*", "*"))
|
|
|
|
.as(Encoders.bean(Relation.class));
|
|
|
|
|
|
|
|
Dataset<Tuple2<String, String>> mergedIds = mergeRels
|
2021-03-31 17:07:13 +02:00
|
|
|
.where(col("relClass").equalTo(ModelConstants.MERGES))
|
2020-04-27 14:52:31 +02:00
|
|
|
.select(col("source"), col("target"))
|
|
|
|
.distinct()
|
|
|
|
.map(
|
|
|
|
(MapFunction<Row, Tuple2<String, String>>) r -> new Tuple2<>(r.getString(1), r.getString(0)),
|
|
|
|
Encoders.tuple(Encoders.STRING(), Encoders.STRING()))
|
|
|
|
.cache();
|
|
|
|
|
|
|
|
final String relationPath = DedupUtility.createEntityPath(graphBasePath, "relation");
|
|
|
|
|
|
|
|
Dataset<Relation> rels = spark.read().textFile(relationPath).map(patchRelFn(), Encoders.bean(Relation.class));
|
|
|
|
|
2021-04-16 16:40:42 +02:00
|
|
|
Dataset<Relation> newRels = createNewRels(rels, mergedIds, getFixRelFn());
|
2020-04-27 14:52:31 +02:00
|
|
|
|
|
|
|
Dataset<Relation> updated = processDataset(
|
|
|
|
processDataset(rels, mergedIds, FieldType.SOURCE, getDeletedFn()),
|
|
|
|
mergedIds,
|
|
|
|
FieldType.TARGET,
|
|
|
|
getDeletedFn());
|
|
|
|
|
2020-07-13 15:30:57 +02:00
|
|
|
save(
|
2020-07-13 16:46:13 +02:00
|
|
|
distinctRelations(
|
|
|
|
newRels
|
|
|
|
.union(updated)
|
|
|
|
.union(mergeRels)
|
2020-11-25 14:55:32 +01:00
|
|
|
.map((MapFunction<Relation, Relation>) r -> r, Encoders.kryo(Relation.class)))
|
|
|
|
.filter((FilterFunction<Relation>) r -> !Objects.equals(r.getSource(), r.getTarget())),
|
2020-07-13 15:30:57 +02:00
|
|
|
outputRelationPath, SaveMode.Overwrite);
|
2020-07-10 15:09:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
private Dataset<Relation> distinctRelations(Dataset<Relation> rels) {
|
|
|
|
return rels
|
2020-07-13 15:30:57 +02:00
|
|
|
.filter(getRelationFilterFunction())
|
2020-07-13 16:46:13 +02:00
|
|
|
.groupByKey(
|
|
|
|
(MapFunction<Relation, String>) r -> String
|
|
|
|
.join(r.getSource(), r.getTarget(), r.getRelType(), r.getSubRelType(), r.getRelClass()),
|
|
|
|
Encoders.STRING())
|
2020-07-10 19:35:10 +02:00
|
|
|
.agg(new RelationAggregator().toColumn())
|
|
|
|
.map((MapFunction<Tuple2<String, Relation>, Relation>) t -> t._2(), Encoders.bean(Relation.class));
|
2020-07-10 15:09:50 +02:00
|
|
|
}
|
|
|
|
|
2021-04-16 16:40:42 +02:00
|
|
|
private static Dataset<Relation> createNewRels(
|
|
|
|
Dataset<Relation> rels,
|
|
|
|
Dataset<Tuple2<String,String>> mergedIds,
|
|
|
|
MapFunction<Tuple2<Tuple2<Tuple3<String, Relation, String>, Tuple2<String, String>>, Tuple2<String, String>>, Relation> mapRel) {
|
|
|
|
|
|
|
|
Dataset<Tuple3<String, Relation, String>> mapped = rels
|
|
|
|
.map(
|
|
|
|
(MapFunction<Relation, Tuple3<String, Relation, String>>) r -> new Tuple3<>(getId(r, FieldType.SOURCE), r, getId(r, FieldType.TARGET)),
|
|
|
|
Encoders.tuple(Encoders.STRING(), Encoders.kryo(Relation.class), Encoders.STRING()));
|
|
|
|
|
|
|
|
Dataset<Tuple2<Tuple3<String, Relation, String>, Tuple2<String, String>>> relSource = mapped
|
|
|
|
.joinWith(mergedIds, mapped.col("_1").equalTo(mergedIds.col("_1")), "left_outer");
|
|
|
|
|
|
|
|
return relSource
|
|
|
|
.joinWith(mergedIds, relSource.col("_1._3").equalTo(mergedIds.col("_2")), "left_outer")
|
|
|
|
.filter((FilterFunction<Tuple2<Tuple2<Tuple3<String, Relation, String>, Tuple2<String, String>>, Tuple2<String, String>>>) r -> r._2() != null || r._1() != null)
|
|
|
|
.map(mapRel, Encoders.bean(Relation.class))
|
|
|
|
.distinct();
|
|
|
|
}
|
|
|
|
|
2020-04-27 14:52:31 +02:00
|
|
|
private static Dataset<Relation> processDataset(
|
|
|
|
Dataset<Relation> rels,
|
|
|
|
Dataset<Tuple2<String, String>> mergedIds,
|
|
|
|
FieldType type,
|
|
|
|
MapFunction<Tuple2<Tuple2<String, Relation>, Tuple2<String, String>>, Relation> mapFn) {
|
|
|
|
final Dataset<Tuple2<String, Relation>> mapped = rels
|
|
|
|
.map(
|
|
|
|
(MapFunction<Relation, Tuple2<String, Relation>>) r -> new Tuple2<>(getId(r, type), r),
|
|
|
|
Encoders.tuple(Encoders.STRING(), Encoders.kryo(Relation.class)));
|
|
|
|
return mapped
|
|
|
|
.joinWith(mergedIds, mapped.col("_1").equalTo(mergedIds.col("_1")), "left_outer")
|
|
|
|
.map(mapFn, Encoders.bean(Relation.class));
|
|
|
|
}
|
|
|
|
|
2020-07-13 15:30:57 +02:00
|
|
|
private FilterFunction<Relation> getRelationFilterFunction() {
|
|
|
|
return (FilterFunction<Relation>) r -> StringUtils.isNotBlank(r.getSource()) ||
|
|
|
|
StringUtils.isNotBlank(r.getTarget()) ||
|
|
|
|
StringUtils.isNotBlank(r.getRelClass()) ||
|
|
|
|
StringUtils.isNotBlank(r.getSubRelType()) ||
|
|
|
|
StringUtils.isNotBlank(r.getRelClass());
|
|
|
|
}
|
|
|
|
|
2020-04-27 14:52:31 +02:00
|
|
|
private static String getId(Relation r, FieldType type) {
|
|
|
|
switch (type) {
|
|
|
|
case SOURCE:
|
|
|
|
return r.getSource();
|
|
|
|
case TARGET:
|
|
|
|
return r.getTarget();
|
|
|
|
default:
|
|
|
|
throw new IllegalArgumentException("");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-04-16 16:40:42 +02:00
|
|
|
private static MapFunction<Tuple2<Tuple2<Tuple3<String, Relation, String>, Tuple2<String, String>>, Tuple2<String, String>>, Relation> getFixRelFn() {
|
2020-04-27 14:52:31 +02:00
|
|
|
return value -> {
|
2021-04-16 16:40:42 +02:00
|
|
|
|
|
|
|
Relation r = value._1()._1()._2();
|
|
|
|
String newSource = value._1()._2() != null ? value._1()._2()._2() : null;
|
|
|
|
String newTarget = value._2() != null ? value._2()._2() : null;
|
|
|
|
|
|
|
|
if (r.getDataInfo() == null) {
|
|
|
|
r.setDataInfo(new DataInfo());
|
2020-04-27 14:52:31 +02:00
|
|
|
}
|
2021-04-16 16:40:42 +02:00
|
|
|
r.getDataInfo().setDeletedbyinference(false);
|
|
|
|
|
|
|
|
if (newSource != null)
|
|
|
|
r.setSource(newSource);
|
|
|
|
|
|
|
|
if (newTarget != null)
|
|
|
|
r.setTarget(newTarget);
|
|
|
|
|
|
|
|
return r;
|
2020-04-27 14:52:31 +02:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static MapFunction<Tuple2<Tuple2<String, Relation>, Tuple2<String, String>>, Relation> getDeletedFn() {
|
|
|
|
return value -> {
|
|
|
|
if (value._2() != null) {
|
|
|
|
Relation r = value._1()._2();
|
|
|
|
if (r.getDataInfo() == null) {
|
|
|
|
r.setDataInfo(new DataInfo());
|
|
|
|
}
|
|
|
|
r.getDataInfo().setDeletedbyinference(true);
|
|
|
|
return r;
|
|
|
|
}
|
|
|
|
return value._1()._2();
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2020-04-18 12:42:58 +02:00
|
|
|
}
|