WIP: trying to find a way to build the records for the index

This commit is contained in:
Claudio Atzori 2020-01-16 12:02:28 +02:00
parent 7ba586d2e5
commit 97c239ee0d
9 changed files with 265 additions and 73 deletions

View File

@ -26,6 +26,11 @@
<artifactId>commons-lang3</artifactId> <artifactId>commons-lang3</artifactId>
</dependency> </dependency>
<dependency>
<groupId>com.fasterxml.jackson.core</groupId>
<artifactId>jackson-databind</artifactId>
</dependency>
<dependency> <dependency>
<groupId>eu.dnetlib.dhp</groupId> <groupId>eu.dnetlib.dhp</groupId>
<artifactId>dhp-common</artifactId> <artifactId>dhp-common</artifactId>

View File

@ -1,5 +1,8 @@
package eu.dnetlib.dhp.schema.oaf; package eu.dnetlib.dhp.schema.oaf;
import com.fasterxml.jackson.core.JsonProcessingException;
import com.fasterxml.jackson.databind.ObjectMapper;
import java.io.Serializable; import java.io.Serializable;
public abstract class Oaf implements Serializable { public abstract class Oaf implements Serializable {
@ -23,4 +26,14 @@ public abstract class Oaf implements Serializable {
public void setLastupdatetimestamp(Long lastupdatetimestamp) { public void setLastupdatetimestamp(Long lastupdatetimestamp) {
this.lastupdatetimestamp = lastupdatetimestamp; this.lastupdatetimestamp = lastupdatetimestamp;
} }
@Override
public String toString() {
try {
return new ObjectMapper().writeValueAsString(this);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
}
} }

View File

@ -1,3 +1,5 @@
sparkDriverMemory=16G sparkDriverMemory=7G
sparkExecutorMemory=16G sparkExecutorMemory=7G
sparkExecutorMemoryOverhead=5G
hive_db_name=claudio hive_db_name=claudio
sourcePath=/tmp/db_openaireplus_services_beta.export.2019.11.06

View File

@ -0,0 +1,53 @@
package eu.dnetlib.dhp.graph;
import com.fasterxml.jackson.core.JsonProcessingException;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.schema.oaf.Relation;
import java.io.Serializable;
public class EntityRelEntity implements Serializable {
private TypedRow source;
private Relation relation;
private TypedRow target;
public EntityRelEntity(TypedRow source) {
this.source = source;
}
public TypedRow getSource() {
return source;
}
public EntityRelEntity setSource(TypedRow source) {
this.source = source;
return this;
}
public Relation getRelation() {
return relation;
}
public EntityRelEntity setRelation(Relation relation) {
this.relation = relation;
return this;
}
public TypedRow getTarget() {
return target;
}
public EntityRelEntity setTarget(TypedRow target) {
this.target = target;
return this;
}
@Override
public String toString() {
try {
return new ObjectMapper().writeValueAsString(this);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
}
}

View File

@ -0,0 +1,139 @@
package eu.dnetlib.dhp.graph;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.schema.oaf.*;
import org.apache.hadoop.io.Text;
import org.apache.spark.api.java.JavaPairRDD;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.api.java.Optional;
import org.apache.spark.api.java.function.PairFunction;
import org.apache.spark.sql.Encoders;
import org.apache.spark.sql.SaveMode;
import org.apache.spark.sql.SparkSession;
import scala.Tuple2;
import java.io.Serializable;
public class GraphJoiner implements Serializable {
public static final int MAX_RELS = 100;
public void join(final SparkSession spark, final String inputPath, final String hiveDbName, final String outPath) {
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
/*
JavaPairRDD<String, TypedRow> entities = sc.sequenceFile(inputPath + "/publication", Text.class, Text.class)
.map(item -> new ObjectMapper().readValue(item._2().toString(), Publication.class))
.map(oaf -> new TypedRow("publication", oaf))
.mapToPair(toPair());
*/
JavaPairRDD<String, TypedRow> entities = sc.sequenceFile(inputPath + "/datasource", Text.class, Text.class)
.map(item -> new ObjectMapper().readValue(item._2().toString(), Datasource.class))
.map(oaf -> new TypedRow("datasource", oaf))
.mapToPair(toPair())
.union(sc.sequenceFile(inputPath + "/organization", Text.class, Text.class)
.map(item -> new ObjectMapper().readValue(item._2().toString(), Organization.class))
.map(oaf -> new TypedRow("organization", oaf))
.mapToPair(toPair()))
.union(sc.sequenceFile(inputPath + "/project", Text.class, Text.class)
.map(item -> new ObjectMapper().readValue(item._2().toString(), Project.class))
.map(oaf -> new TypedRow("project", oaf))
.mapToPair(toPair()))
.union(sc.sequenceFile(inputPath + "/dataset", Text.class, Text.class)
.map(item -> new ObjectMapper().readValue(item._2().toString(), Dataset.class))
.map(oaf -> new TypedRow("dataset", oaf))
.mapToPair(toPair()))
.union(sc.sequenceFile(inputPath + "/otherresearchproduct", Text.class, Text.class)
.map(item -> new ObjectMapper().readValue(item._2().toString(), OtherResearchProduct.class))
.map(oaf -> new TypedRow("otherresearchproduct", oaf))
.mapToPair(toPair()))
.union(sc.sequenceFile(inputPath + "/software", Text.class, Text.class)
.map(item -> new ObjectMapper().readValue(item._2().toString(), Software.class))
.map(oaf -> new TypedRow("software", oaf))
.mapToPair(toPair()));
/*
.union(sc.sequenceFile(inputPath + "/publication", Text.class, Text.class)
.map(item -> new ObjectMapper().readValue(item._2().toString(), Publication.class))
.map(oaf -> new TypedRow("publication", oaf))
.mapToPair(toPair()));
*/
/*
JavaRDD<Relation> rels = sc.sequenceFile(inputPath + "/relation", Text.class, Text.class)
.map(item -> new ObjectMapper().readValue(item._2().toString(), Relation.class))
.map(oaf -> new TypedRow("relation", oaf))
.mapToPair(toPair())
.groupByKey()
.map(t -> Iterables.limit(t._2(), MAX_RELS))
.flatMap(t -> t.iterator())
.map(t -> (Relation) t.getOaf());
spark.createDataset(rels.rdd(), Encoders.bean(Relation.class))
.write()
.mode(SaveMode.Overwrite)
.saveAsTable(hiveDbName + ".relation_100");
*/
JavaPairRDD<String, TypedRow> bounded_rels = spark.table(hiveDbName + ".relation_" + MAX_RELS)
.as(Encoders.bean(Relation.class))
.javaRDD()
.map(r -> new TypedRow("relation", r))
.mapToPair(toPair());
// build the adjacency list: e -> r
JavaPairRDD<String, Tuple2<TypedRow, Optional<TypedRow>>> adjacency_list = entities.leftOuterJoin(bounded_rels);
JavaRDD<EntityRelEntity> linked_entities = adjacency_list
.mapToPair(toPairTarget()) // make rel.targetid explicit so that we can join it
.leftOuterJoin(entities) // again with the entities to get the target entity
.map(l -> toEntityRelEntity(l)); // and map it to a more readable representation
spark.createDataFrame(linked_entities, EntityRelEntity.class)
.write()
.mode(SaveMode.Overwrite)
.saveAsTable(hiveDbName + ".linked_entities");
}
private EntityRelEntity toEntityRelEntity(Tuple2<String, Tuple2<Tuple2<String, Tuple2<TypedRow, Optional<TypedRow>>>, Optional<TypedRow>>> l) {
// extract the entity source
final EntityRelEntity res = new EntityRelEntity(l._2()._1()._2()._1());
if(l._2()._1()._2()._2().isPresent() && l._2()._2().isPresent()) {
// extract the relationship
res.setRelation((Relation) l._2()._1()._2()._2().get().getOaf());
// extract the related entity
res.setTarget(l._2()._2().get());
}
return res;
}
private PairFunction<Tuple2<String, Tuple2<TypedRow, Optional<TypedRow>>>, String, Tuple2<String, Tuple2<TypedRow, Optional<TypedRow>>>> toPairTarget() {
return e -> {
Optional<TypedRow> o = e._2()._2();
if (o.isPresent()) {
return new Tuple2<>(((Relation) o.get().getOaf()).getTarget(), e);
} else {
return new Tuple2<>(null, e);
}
};
}
private PairFunction<TypedRow, String, TypedRow> toPair() {
return e -> {
if (!"relation".equals(e.getType())) {
return new Tuple2<>( ((OafEntity) e.getOaf()).getId(), e);
} else {
return new Tuple2<>( ((Relation) e.getOaf()).getSource(), e);
}
};
}
}

View File

@ -1,32 +1,14 @@
package eu.dnetlib.dhp.graph; package eu.dnetlib.dhp.graph;
import com.google.common.collect.Sets;
import eu.dnetlib.dhp.application.ArgumentApplicationParser; import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.schema.common.EntityPayload;
import org.apache.commons.io.IOUtils; import org.apache.commons.io.IOUtils;
import org.apache.hadoop.fs.FileSystem; import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path; import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.compress.GzipCodec;
import org.apache.spark.api.java.function.MapFunction;
import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Row;
import org.apache.spark.sql.SaveMode;
import org.apache.spark.sql.SparkSession; import org.apache.spark.sql.SparkSession;
import scala.Tuple2;
import scala.runtime.AbstractFunction1;
import java.util.List;
import java.util.Set;
import java.util.stream.Collectors;
import static org.apache.commons.lang3.StringUtils.substringAfter;
import static org.apache.commons.lang3.StringUtils.substringBefore;
import static org.apache.spark.sql.Encoders.bean;
public class SparkGraphIndexingJob { public class SparkGraphIndexingJob {
private final static String ENTITY_NODES_PATH = "/tmp/entity_node"; private final static String ENTITY_NODES_PATH = "/tmp/entity_node";
private static final long LIMIT = 100;
public static void main(String[] args) throws Exception { public static void main(String[] args) throws Exception {
@ -37,13 +19,10 @@ public class SparkGraphIndexingJob {
.appName(SparkGraphIndexingJob.class.getSimpleName()) .appName(SparkGraphIndexingJob.class.getSimpleName())
.master(parser.get("master")) .master(parser.get("master"))
.config("hive.metastore.uris", parser.get("hive_metastore_uris")) .config("hive.metastore.uris", parser.get("hive_metastore_uris"))
.config("spark.driver.cores", 1)
.config("spark.executor.cores", 1)
.config("spark.yarn.executor.memoryOverhead", "4G")
.config("spark.yarn.driver.memoryOverhead", "4G")
.enableHiveSupport() .enableHiveSupport()
.getOrCreate(); .getOrCreate();
final String inputPath = parser.get("sourcePath");
final String hiveDbName = parser.get("hive_db_name"); final String hiveDbName = parser.get("hive_db_name");
final FileSystem fs = FileSystem.get(spark.sparkContext().hadoopConfiguration()); final FileSystem fs = FileSystem.get(spark.sparkContext().hadoopConfiguration());
@ -51,52 +30,7 @@ public class SparkGraphIndexingJob {
fs.delete(new Path(ENTITY_NODES_PATH), true); fs.delete(new Path(ENTITY_NODES_PATH), true);
} }
spark new GraphJoiner().join(spark, inputPath, hiveDbName, ENTITY_NODES_PATH);
.sql(getJoinEntitiesSQL(hiveDbName))
.transform(toEntityNode())
/*
.map((MapFunction<EntityNode, String>) r -> {
return null;
}, bean(String.class))
*/
.rdd()
.saveAsTextFile(ENTITY_NODES_PATH, GzipCodec.class);
}
private static AbstractFunction1<Dataset<Row>, Dataset<EntityNode>> toEntityNode() {
return new AbstractFunction1<Dataset<Row>, Dataset<EntityNode>>() {
@Override
public Dataset<EntityNode> apply(Dataset<Row> d) {
return d.map((MapFunction<Row, EntityNode>) r -> {
final List<String> res = r.getList(r.fieldIndex("related_entity"));
final byte[] payload = r.getAs("payload");
return new EntityNode(r.getAs("id"), r.getAs("type"), new String(payload))
.setRelatedEntities(res
.stream()
.map(re -> new Tuple2<>(substringBefore(re, "@@"), substringAfter(re, "@@")))
.map(re -> new RelatedEntity(r.getAs("reltype"), r.getAs("subreltype"), r.getAs("relclass"), re._1(), re._2()))
.limit(LIMIT)
.collect(Collectors.toList()));
}, bean(EntityNode.class));
}
};
}
private static String getJoinEntitiesSQL(String hiveDbName) {
return String.format(
"SELECT " +
"E_s.id AS id, " +
"E_s.type AS type, " +
"E_s.payload AS payload, " +
"r.reltype AS reltype, r.subreltype AS subreltype, r.relclass AS relclass, " +
"collect_list(concat(E_t.type, '@@', E_t.payload)) AS related_entity " +
"FROM %s.entities " + "" /*"TABLESAMPLE(0.1 PERCENT) "*/ + "E_s " +
"LEFT JOIN %s.relation r ON (r.source = E_s.id) " +
"JOIN %s.entities E_t ON (E_t.id = r.target) \n" +
"GROUP BY E_s.id, E_s.type, E_s.payload, r.reltype, r.subreltype, r.relclass", hiveDbName, hiveDbName, hiveDbName);
} }
} }

View File

@ -0,0 +1,44 @@
package eu.dnetlib.dhp.graph;
import com.fasterxml.jackson.core.JsonProcessingException;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.schema.oaf.Oaf;
import java.io.Serializable;
public class TypedRow implements Serializable {
private String type;
private Oaf oaf;
public TypedRow(String type, Oaf oaf) {
this.type = type;
this.oaf = oaf;
}
public String getType() {
return type;
}
public TypedRow setType(String type) {
this.type = type;
return this;
}
public Oaf getOaf() {
return oaf;
}
public TypedRow setOaf(Oaf oaf) {
this.oaf = oaf;
return this;
}
@Override
public String toString() {
try {
return new ObjectMapper().writeValueAsString(this);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
}
}

View File

@ -1,5 +1,6 @@
[ [
{"paramName":"mt", "paramLongName":"master", "paramDescription": "should be local or yarn", "paramRequired": true}, {"paramName":"mt", "paramLongName":"master", "paramDescription": "should be local or yarn", "paramRequired": true},
{"paramName":"h", "paramLongName":"hive_metastore_uris","paramDescription": "the hive metastore uris", "paramRequired": true}, {"paramName":"h", "paramLongName":"hive_metastore_uris","paramDescription": "the hive metastore uris", "paramRequired": true},
{"paramName":"db", "paramLongName":"hive_db_name", "paramDescription": "the target hive database name", "paramRequired": true} {"paramName":"db", "paramLongName":"hive_db_name", "paramDescription": "the target hive database name", "paramRequired": true},
{"paramName":"s", "paramLongName":"sourcePath", "paramDescription": "the path of the sequencial file to read", "paramRequired": true}
] ]

View File

@ -33,8 +33,9 @@
<name>GraphIndexing</name> <name>GraphIndexing</name>
<class>eu.dnetlib.dhp.graph.SparkGraphIndexingJob</class> <class>eu.dnetlib.dhp.graph.SparkGraphIndexingJob</class>
<jar>dhp-graph-provision-${projectVersion}.jar</jar> <jar>dhp-graph-provision-${projectVersion}.jar</jar>
<spark-opts>--executor-memory ${sparkExecutorMemory} --executor-cores ${sparkExecutorCores} --driver-memory=${sparkDriverMemory} --conf spark.extraListeners="com.cloudera.spark.lineage.NavigatorAppListener" --conf spark.sql.queryExecutionListeners="com.cloudera.spark.lineage.NavigatorQueryListener" --conf spark.sql.warehouse.dir="/user/hive/warehouse"</spark-opts> <spark-opts>--executor-memory ${sparkExecutorMemory} --executor-cores ${sparkExecutorCores} --driver-memory=${sparkDriverMemory} --conf spark.extraListeners="com.cloudera.spark.lineage.NavigatorAppListener" --conf spark.sql.queryExecutionListeners="com.cloudera.spark.lineage.NavigatorQueryListener" --conf spark.sql.warehouse.dir="/user/hive/warehouse" --conf spark.yarn.executor.memoryOverhead=${sparkExecutorMemoryOverhead}</spark-opts>
<arg>-mt</arg> <arg>yarn-cluster</arg> <arg>-mt</arg> <arg>yarn-cluster</arg>
<arg>--sourcePath</arg><arg>${sourcePath}</arg>
<arg>--hive_db_name</arg><arg>${hive_db_name}</arg> <arg>--hive_db_name</arg><arg>${hive_db_name}</arg>
<arg>--hive_metastore_uris</arg><arg>${hive_metastore_uris}</arg> <arg>--hive_metastore_uris</arg><arg>${hive_metastore_uris}</arg>
</spark> </spark>