D-Net
/
dnet-hadoop
13
4
Fork 10
Code Issues 14 Pull Requests 7 Projects 2 Releases 1 Wiki Activity

Fix ensure all relations are written out, not only those managed by dedup

This commit is contained in:
Giambattista Bloisi 2023-08-28 15:36:02 +02:00
parent 0d7b2bf83d
commit a860e19423
1 changed files with 38 additions and 52 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

90
dhp-workflows/dhp-dedup-openaire/src/main/java/eu/dnetlib/dhp/oa/dedup/SparkPropagateRelation.java
View File

@ -8,6 +8,7 @@ import eu.dnetlib.dhp.schema.oaf.DataInfo;
import eu.dnetlib.dhp.schema.oaf.Relation; import eu.dnetlib.dhp.schema.oaf.Relation;
import eu.dnetlib.dhp.utils.ISLookupClientFactory; import eu.dnetlib.dhp.utils.ISLookupClientFactory;
import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpService; import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpService;
import org.apache.commons.beanutils.BeanUtils;
import org.apache.commons.io.IOUtils; import org.apache.commons.io.IOUtils;
import org.apache.commons.lang3.StringUtils; import org.apache.commons.lang3.StringUtils;
import org.apache.spark.SparkConf; import org.apache.spark.SparkConf;
@ -20,6 +21,9 @@ import org.slf4j.LoggerFactory;
import scala.Tuple2; import scala.Tuple2;
import scala.Tuple3; import scala.Tuple3;
import java.util.Arrays;
import java.util.Collections;
import java.util.Iterator;
import java.util.Objects; import java.util.Objects;
import static org.apache.spark.sql.functions.col; import static org.apache.spark.sql.functions.col;
@ -80,68 +84,51 @@ public class SparkPropagateRelation extends AbstractSparkAction {
.distinct() .distinct()
.cache(); .cache();
final String inputRelationPath = DedupUtility.createEntityPath(graphBasePath, "relation"); Dataset<Row> allRels = spark.read()
.schema(REL_BEAN_ENC.schema())
.json(DedupUtility.createEntityPath(graphBasePath, "relation"));
Dataset<Relation> rels = spark.read().schema(REL_BEAN_ENC.schema()).json(inputRelationPath) Dataset<Relation> dedupedRels = allRels
.as(REL_BEAN_ENC) .joinWith(mergedIds, allRels.col("source").equalTo(mergedIds.col("mergedObjectID")), "left_outer")
// .map((MapFunction<Relation, Relation>) rel -> {
// if (rel.getDataInfo() == null) {
// rel.setDataInfo(new DataInfo());
// }
// return rel;
// }, REL_BEAN_ENC)
;
Dataset<Tuple3<Relation, String, String>> dedupedRels = rels
.joinWith(mergedIds, rels.col("source").equalTo(mergedIds.col("mergedObjectID")), "left_outer")
.joinWith(mergedIds, col("_1.target").equalTo(mergedIds.col("mergedObjectID")), "left_outer") .joinWith(mergedIds, col("_1.target").equalTo(mergedIds.col("mergedObjectID")), "left_outer")
.filter("_1._2 IS NOT NULL OR _2 IS NOT NULL")
.select("_1._1", "_1._2.dedupID", "_2.dedupID") .select("_1._1", "_1._2.dedupID", "_2.dedupID")
.as(Encoders.tuple(REL_BEAN_ENC, Encoders.STRING(), Encoders.STRING())) .as(Encoders.tuple(REL_BEAN_ENC, Encoders.STRING(), Encoders.STRING()))
.cache(); .flatMap(SparkPropagateRelation::addInferredRelations, REL_KRYO_ENC);
mergedIds.unpersist(); Dataset<Relation> processedRelations = distinctRelations(dedupedRels.union(mergeRels.map((MapFunction<Relation, Relation>) r -> r, REL_KRYO_ENC)))
.filter((FilterFunction<Relation>) r -> !Objects.equals(r.getSource(), r.getTarget()));
Dataset<Relation> newRels = dedupedRels save(processedRelations, outputRelationPath, SaveMode.Overwrite);
.map((MapFunction<Tuple3<Relation, String, String>, Relation>) t -> { }
Relation r = t._1();
String newSource = t._2();
String newTarget = t._3();
if (r.getDataInfo() == null) { private static Iterator<Relation> addInferredRelations(Tuple3<Relation, String, String> t) throws Exception {
r.setDataInfo(new DataInfo()); Relation existingRel = t._1();
} String newSource = t._2();
r.getDataInfo().setDeletedbyinference(false); String newTarget = t._3();
if (newSource != null) if (newSource == null && newTarget == null) {
r.setSource(newSource); return Collections.singleton(t._1()).iterator();
}
if (newTarget != null) // update existing relation
r.setTarget(newTarget); if (existingRel.getDataInfo() == null) {
existingRel.setDataInfo(new DataInfo());
}
existingRel.getDataInfo().setDeletedbyinference(true);
return r; // Create new relation inferred by dedupIDs
}, REL_BEAN_ENC) Relation inferredRel = (Relation) BeanUtils.cloneBean(existingRel);
.distinct();
Dataset<Relation> updated = dedupedRels inferredRel.setDataInfo((DataInfo) BeanUtils.cloneBean(existingRel.getDataInfo()));
.map((MapFunction<Tuple3<Relation, String, String>, Relation>) t -> { inferredRel.getDataInfo().setDeletedbyinference(false);
Relation r = t._1();
if (r.getDataInfo() == null) {
r.setDataInfo(new DataInfo());
}
r.getDataInfo().setDeletedbyinference(true);
return r;
}, REL_BEAN_ENC);
save( if (newSource != null)
distinctRelations( inferredRel.setSource(newSource);
newRels
.union(updated) if (newTarget != null)
.union(mergeRels) inferredRel.setTarget(newTarget);
.map((MapFunction<Relation, Relation>) r -> r, REL_KRYO_ENC)
) return Arrays.asList(existingRel, inferredRel).iterator();
.filter((FilterFunction<Relation>) r -> !Objects.equals(r.getSource(), r.getTarget())),
outputRelationPath, SaveMode.Overwrite);
} }
private Dataset<Relation> distinctRelations(Dataset<Relation> rels) { private Dataset<Relation> distinctRelations(Dataset<Relation> rels) {
@ -156,8 +143,7 @@ public class SparkPropagateRelation extends AbstractSparkAction {
return b; return b;
} }
) )
.map((MapFunction<Tuple2<String, Relation>, Relation>) Tuple2::_2, REL_BEAN_ENC);
.map((MapFunction<Tuple2<String, Relation>, Relation>) Tuple2::_2, REL_BEAN_ENC);
} }
private FilterFunction<Relation> getRelationFilterFunction() { private FilterFunction<Relation> getRelationFilterFunction() {