323 lines
12 KiB
Java
323 lines
12 KiB
Java
|
|
package eu.dnetlib.dhp.oa.dedup;
|
|
|
|
import java.io.IOException;
|
|
import java.util.*;
|
|
import java.util.stream.Collectors;
|
|
import java.util.stream.StreamSupport;
|
|
|
|
import org.apache.commons.io.IOUtils;
|
|
import org.apache.spark.SparkConf;
|
|
import org.apache.spark.api.java.JavaRDD;
|
|
import org.apache.spark.api.java.function.MapFunction;
|
|
import org.apache.spark.sql.Dataset;
|
|
import org.apache.spark.sql.Encoders;
|
|
import org.apache.spark.sql.SaveMode;
|
|
import org.apache.spark.sql.SparkSession;
|
|
import org.slf4j.Logger;
|
|
import org.slf4j.LoggerFactory;
|
|
|
|
import com.google.common.collect.Lists;
|
|
|
|
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
|
|
import eu.dnetlib.dhp.oa.dedup.model.OrgSimRel;
|
|
import eu.dnetlib.dhp.schema.common.ModelConstants;
|
|
import eu.dnetlib.dhp.schema.common.ModelSupport;
|
|
import eu.dnetlib.dhp.schema.oaf.*;
|
|
import eu.dnetlib.dhp.utils.ISLookupClientFactory;
|
|
import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpService;
|
|
import scala.Tuple2;
|
|
import scala.Tuple3;
|
|
|
|
public class SparkPrepareOrgRels extends AbstractSparkAction {
|
|
|
|
private static final Logger log = LoggerFactory.getLogger(SparkPrepareOrgRels.class);
|
|
|
|
public SparkPrepareOrgRels(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/prepareOrgRels_parameters.json")));
|
|
parser.parseArgument(args);
|
|
|
|
SparkConf conf = new SparkConf();
|
|
conf.set("spark.serializer", "org.apache.spark.serializer.KryoSerializer");
|
|
conf.registerKryoClasses(ModelSupport.getOafModelClasses());
|
|
|
|
new SparkPrepareOrgRels(parser, getSparkSession(conf))
|
|
.run(ISLookupClientFactory.getLookUpService(parser.get("isLookUpUrl")));
|
|
}
|
|
|
|
@Override
|
|
public void run(ISLookUpService isLookUpService) throws IOException {
|
|
|
|
final String graphBasePath = parser.get("graphBasePath");
|
|
final String isLookUpUrl = parser.get("isLookUpUrl");
|
|
final String actionSetId = parser.get("actionSetId");
|
|
final String workingPath = parser.get("workingPath");
|
|
final int numConnections = Optional
|
|
.ofNullable(parser.get("numConnections"))
|
|
.map(Integer::valueOf)
|
|
.orElse(NUM_CONNECTIONS);
|
|
|
|
final String dbUrl = parser.get("dbUrl");
|
|
final String dbTable = parser.get("dbTable");
|
|
final String dbUser = parser.get("dbUser");
|
|
final String dbPwd = parser.get("dbPwd");
|
|
|
|
log.info("graphBasePath: '{}'", graphBasePath);
|
|
log.info("isLookUpUrl: '{}'", isLookUpUrl);
|
|
log.info("actionSetId: '{}'", actionSetId);
|
|
log.info("workingPath: '{}'", workingPath);
|
|
log.info("numPartitions: '{}'", numConnections);
|
|
log.info("dbUrl: '{}'", dbUrl);
|
|
log.info("dbUser: '{}'", dbUser);
|
|
log.info("table: '{}'", dbTable);
|
|
log.info("dbPwd: '{}'", "xxx");
|
|
|
|
final String mergeRelPath = DedupUtility.createMergeRelPath(workingPath, actionSetId, "organization");
|
|
final String entityPath = DedupUtility.createEntityPath(graphBasePath, "organization");
|
|
final String relationPath = DedupUtility.createEntityPath(graphBasePath, "relation");
|
|
|
|
Dataset<OrgSimRel> relations = createRelations(spark, mergeRelPath, relationPath, entityPath);
|
|
|
|
final Properties connectionProperties = new Properties();
|
|
connectionProperties.put("user", dbUser);
|
|
connectionProperties.put("password", dbPwd);
|
|
|
|
relations
|
|
.repartition(numConnections)
|
|
.write()
|
|
.mode(SaveMode.Overwrite)
|
|
.jdbc(dbUrl, dbTable, connectionProperties);
|
|
|
|
}
|
|
|
|
private static boolean filterRels(Relation rel, String entityType) {
|
|
|
|
switch (entityType) {
|
|
case "result":
|
|
if (rel.getRelClass().equals(ModelConstants.IS_DIFFERENT_FROM)
|
|
&& rel.getRelType().equals(ModelConstants.RESULT_RESULT)
|
|
&& rel.getSubRelType().equals(ModelConstants.DEDUP))
|
|
return true;
|
|
break;
|
|
case "organization":
|
|
if (rel.getRelClass().equals(ModelConstants.IS_DIFFERENT_FROM)
|
|
&& rel.getRelType().equals(ModelConstants.ORG_ORG_RELTYPE)
|
|
&& rel.getSubRelType().equals(ModelConstants.DEDUP))
|
|
return true;
|
|
break;
|
|
default:
|
|
return false;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
// create openorgs simrels <best id, other id> starting from mergerels, remove the diffrels
|
|
public static Dataset<OrgSimRel> createRelations(
|
|
final SparkSession spark,
|
|
final String mergeRelsPath,
|
|
final String relationPath,
|
|
final String entitiesPath) {
|
|
|
|
// collect diffrels from the raw graph relations: <<best id, other id>, "diffRel">
|
|
JavaRDD<Tuple2<Tuple2<String, String>, String>> diffRels = spark
|
|
.read()
|
|
.textFile(relationPath)
|
|
.map(patchRelFn(), Encoders.bean(Relation.class))
|
|
.toJavaRDD()
|
|
.filter(r -> filterRels(r, "organization"))
|
|
// put the best id as source of the diffrel: <best id, other id>
|
|
.map(rel -> {
|
|
if (DedupUtility.compareOpenOrgIds(rel.getSource(), rel.getTarget()) < 0)
|
|
return new Tuple2<>(new Tuple2<>(rel.getSource(), rel.getTarget()), "diffRel");
|
|
else
|
|
return new Tuple2<>(new Tuple2<>(rel.getTarget(), rel.getSource()), "diffRel");
|
|
})
|
|
.distinct();
|
|
log.info("Number of DiffRels collected: {}", diffRels.count());
|
|
|
|
// collect all the organizations
|
|
Dataset<Tuple2<String, Organization>> entities = spark
|
|
.read()
|
|
.textFile(entitiesPath)
|
|
.map(
|
|
(MapFunction<String, Tuple2<String, Organization>>) it -> {
|
|
Organization entity = OBJECT_MAPPER.readValue(it, Organization.class);
|
|
return new Tuple2<>(entity.getId(), entity);
|
|
},
|
|
Encoders.tuple(Encoders.STRING(), Encoders.kryo(Organization.class)));
|
|
|
|
// relations with their group (connected component id)
|
|
JavaRDD<Tuple2<Tuple2<String, String>, String>> rawOpenorgsRels = spark
|
|
.read()
|
|
.load(mergeRelsPath)
|
|
.as(Encoders.bean(Relation.class))
|
|
.where("relClass == 'merges'")
|
|
.toJavaRDD()
|
|
.mapToPair(r -> new Tuple2<>(r.getSource(), r.getTarget()))
|
|
.filter(t -> !t._2().contains("openorgsmesh")) // remove openorgsmesh: they are only for dedup
|
|
.groupByKey()
|
|
.map(g -> Lists.newArrayList(g._2()))
|
|
.filter(l -> l.size() > 1)
|
|
.flatMap(l -> {
|
|
String groupId = "group::" + UUID.randomUUID();
|
|
List<String> ids = sortIds(l); // sort IDs by type
|
|
List<Tuple2<Tuple2<String, String>, String>> rels = new ArrayList<>();
|
|
String source = ids.get(0);
|
|
for (String target : ids) {
|
|
rels.add(new Tuple2<>(new Tuple2<>(source, target), groupId));
|
|
}
|
|
|
|
return rels.iterator();
|
|
});
|
|
log.info("Number of Raw Openorgs Relations created: {}", rawOpenorgsRels.count());
|
|
|
|
// filter out diffRels
|
|
JavaRDD<Tuple3<String, String, String>> openorgsRels = rawOpenorgsRels
|
|
.union(diffRels)
|
|
// concatenation of source and target: <source|||target, group id> or <source|||target, "diffRel">
|
|
.mapToPair(t -> new Tuple2<>(t._1()._1() + "@@@" + t._1()._2(), t._2()))
|
|
.groupByKey()
|
|
.map(
|
|
g -> new Tuple2<>(g._1(), StreamSupport
|
|
.stream(g._2().spliterator(), false)
|
|
.collect(Collectors.toList())))
|
|
// <source|||target, list(group_id, "diffRel")>: take only relations with only the group_id, it
|
|
// means they are correct. If the diffRel is present the relation has to be removed
|
|
.filter(g -> g._2().size() == 1 && g._2().get(0).contains("group::"))
|
|
.map(
|
|
t -> new Tuple3<>(
|
|
t._1().split("@@@")[0],
|
|
t._1().split("@@@")[1],
|
|
t._2().get(0)));
|
|
log.info("Number of Openorgs Relations created: '{}'", openorgsRels.count());
|
|
|
|
// <best ID basing on priority, ID, groupID>
|
|
Dataset<Tuple3<String, String, String>> relations = spark
|
|
.createDataset(
|
|
openorgsRels.rdd(),
|
|
Encoders.tuple(Encoders.STRING(), Encoders.STRING(), Encoders.STRING()));
|
|
|
|
// create orgsimrels
|
|
Dataset<Tuple2<String, OrgSimRel>> relations2 = relations
|
|
.joinWith(entities, relations.col("_2").equalTo(entities.col("_1")), "inner")
|
|
.map(
|
|
(MapFunction<Tuple2<Tuple3<String, String, String>, Tuple2<String, Organization>>, OrgSimRel>) r -> {
|
|
final Organization o = r._2()._2();
|
|
return new OrgSimRel(
|
|
r._1()._1(),
|
|
o.getOriginalId().get(0),
|
|
Optional.ofNullable(o.getLegalname()).map(Field::getValue).orElse(""),
|
|
Optional.ofNullable(o.getLegalshortname()).map(Field::getValue).orElse(""),
|
|
Optional.ofNullable(o.getCountry()).map(Qualifier::getClassid).orElse(""),
|
|
Optional.ofNullable(o.getWebsiteurl()).map(Field::getValue).orElse(""),
|
|
Optional
|
|
.ofNullable(o.getCollectedfrom())
|
|
.map(c -> Optional.ofNullable(c.get(0)).map(KeyValue::getValue).orElse(""))
|
|
.orElse(""),
|
|
r._1()._3(),
|
|
structuredPropertyListToString(o.getPid()));
|
|
},
|
|
Encoders.bean(OrgSimRel.class))
|
|
.map(
|
|
(MapFunction<OrgSimRel, Tuple2<String, OrgSimRel>>) o -> new Tuple2<>(o.getLocal_id(), o),
|
|
Encoders.tuple(Encoders.STRING(), Encoders.bean(OrgSimRel.class)));
|
|
|
|
return relations2
|
|
.joinWith(entities, relations2.col("_1").equalTo(entities.col("_1")), "inner")
|
|
.map(
|
|
(MapFunction<Tuple2<Tuple2<String, OrgSimRel>, Tuple2<String, Organization>>, OrgSimRel>) r -> {
|
|
OrgSimRel orgSimRel = r._1()._2();
|
|
orgSimRel.setLocal_id(r._2()._2().getOriginalId().get(0));
|
|
return orgSimRel;
|
|
},
|
|
Encoders.bean(OrgSimRel.class));
|
|
|
|
}
|
|
|
|
// Sort IDs basing on the type. Priority: 1) openorgs, 2)corda, 3)alphabetic
|
|
public static List<String> sortIds(List<String> ids) {
|
|
ids.sort((o1, o2) -> DedupUtility.compareOpenOrgIds(o1, o2));
|
|
return ids;
|
|
}
|
|
|
|
public static Dataset<OrgSimRel> createRelationsFromScratch(
|
|
final SparkSession spark,
|
|
final String mergeRelsPath,
|
|
final String entitiesPath) {
|
|
|
|
// <id, json_entity>
|
|
Dataset<Tuple2<String, Organization>> entities = spark
|
|
.read()
|
|
.textFile(entitiesPath)
|
|
.map(
|
|
(MapFunction<String, Tuple2<String, Organization>>) it -> {
|
|
Organization entity = OBJECT_MAPPER.readValue(it, Organization.class);
|
|
return new Tuple2<>(entity.getId(), entity);
|
|
},
|
|
Encoders.tuple(Encoders.STRING(), Encoders.kryo(Organization.class)));
|
|
|
|
Dataset<Tuple2<String, String>> relations = spark
|
|
.createDataset(
|
|
spark
|
|
.read()
|
|
.load(mergeRelsPath)
|
|
.as(Encoders.bean(Relation.class))
|
|
.where("relClass == 'merges'")
|
|
.toJavaRDD()
|
|
.mapToPair(r -> new Tuple2<>(r.getSource(), r.getTarget()))
|
|
.groupByKey()
|
|
.flatMap(g -> {
|
|
List<Tuple2<String, String>> rels = new ArrayList<>();
|
|
for (String id1 : g._2()) {
|
|
for (String id2 : g._2()) {
|
|
if (!id1.equals(id2))
|
|
if (id1.contains(DedupUtility.OPENORGS_ID_PREFIX) && !id2.contains("openorgsmesh"))
|
|
rels.add(new Tuple2<>(id1, id2));
|
|
}
|
|
}
|
|
return rels.iterator();
|
|
})
|
|
.rdd(),
|
|
Encoders.tuple(Encoders.STRING(), Encoders.STRING()));
|
|
|
|
Dataset<Tuple2<String, OrgSimRel>> relations2 = relations // <openorgs, corda>
|
|
.joinWith(entities, relations.col("_2").equalTo(entities.col("_1")), "inner")
|
|
.map(
|
|
(MapFunction<Tuple2<Tuple2<String, String>, Tuple2<String, Organization>>, OrgSimRel>) r -> new OrgSimRel(
|
|
r._1()._1(),
|
|
r._2()._2().getOriginalId().get(0),
|
|
r._2()._2().getLegalname() != null ? r._2()._2().getLegalname().getValue() : "",
|
|
r._2()._2().getLegalshortname() != null ? r._2()._2().getLegalshortname().getValue() : "",
|
|
r._2()._2().getCountry() != null ? r._2()._2().getCountry().getClassid() : "",
|
|
r._2()._2().getWebsiteurl() != null ? r._2()._2().getWebsiteurl().getValue() : "",
|
|
r._2()._2().getCollectedfrom().get(0).getValue(),
|
|
"group::" + r._1()._1(),
|
|
structuredPropertyListToString(r._2()._2().getPid())),
|
|
Encoders.bean(OrgSimRel.class))
|
|
.map(
|
|
(MapFunction<OrgSimRel, Tuple2<String, OrgSimRel>>) o -> new Tuple2<>(o.getLocal_id(), o),
|
|
Encoders.tuple(Encoders.STRING(), Encoders.bean(OrgSimRel.class)));
|
|
|
|
return relations2
|
|
.joinWith(entities, relations2.col("_1").equalTo(entities.col("_1")), "inner")
|
|
.map(
|
|
(MapFunction<Tuple2<Tuple2<String, OrgSimRel>, Tuple2<String, Organization>>, OrgSimRel>) r -> {
|
|
OrgSimRel orgSimRel = r._1()._2();
|
|
orgSimRel.setLocal_id(r._2()._2().getOriginalId().get(0));
|
|
return orgSimRel;
|
|
},
|
|
Encoders.bean(OrgSimRel.class));
|
|
|
|
}
|
|
|
|
}
|