Changed the way to find results linked to projects. We verify to actually have the project on the graph before selecting the result

This commit is contained in:
Miriam Baglioni 2021-07-13 18:13:32 +02:00
parent 52ce35d57b
commit 320cf02d96
1 changed files with 17 additions and 18 deletions

View File

@ -24,7 +24,8 @@ import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.oaf.Relation; import eu.dnetlib.dhp.schema.oaf.Relation;
import eu.dnetlib.dhp.schema.oaf.Result; import eu.dnetlib.dhp.schema.oaf.Result;
import scala.Tuple2; import scala.Tuple2;
import org.apache.spark.sql.*;
import eu.dnetlib.dhp.schema.oaf.Project;
/** /**
* Selects the results linked to projects. Only for these results the dump will be performed. * Selects the results linked to projects. Only for these results the dump will be performed.
* The code to perform the dump and to expend the dumped results with the informaiton related to projects * The code to perform the dump and to expend the dumped results with the informaiton related to projects
@ -59,8 +60,8 @@ public class SparkResultLinkedToProject implements Serializable {
final String resultClassName = parser.get("resultTableName"); final String resultClassName = parser.get("resultTableName");
log.info("resultTableName: {}", resultClassName); log.info("resultTableName: {}", resultClassName);
final String relationPath = parser.get("relationPath"); final String graphPath = parser.get("graphPath");
log.info("relationPath: {}", relationPath); log.info("graphPath: {}", graphPath);
Class<? extends Result> inputClazz = (Class<? extends Result>) Class.forName(resultClassName); Class<? extends Result> inputClazz = (Class<? extends Result>) Class.forName(resultClassName);
SparkConf conf = new SparkConf(); SparkConf conf = new SparkConf();
@ -70,34 +71,32 @@ public class SparkResultLinkedToProject implements Serializable {
isSparkSessionManaged, isSparkSessionManaged,
spark -> { spark -> {
Utils.removeOutputDir(spark, outputPath); Utils.removeOutputDir(spark, outputPath);
writeResultsLinkedToProjects(spark, inputClazz, inputPath, outputPath, relationPath); writeResultsLinkedToProjects(spark, inputClazz, inputPath, outputPath, graphPath);
}); });
} }
private static <R extends Result> void writeResultsLinkedToProjects(SparkSession spark, Class<R> inputClazz, private static <R extends Result> void writeResultsLinkedToProjects(SparkSession spark, Class<R> inputClazz,
String inputPath, String outputPath, String relationPath) { String inputPath, String outputPath, String graphPath) {
Dataset<R> results = Utils Dataset<R> results = Utils
.readPath(spark, inputPath, inputClazz) .readPath(spark, inputPath, inputClazz)
.filter("dataInfo.deletedbyinference = false and datainfo.invisible = false"); .filter("dataInfo.deletedbyinference = false and datainfo.invisible = false");
Dataset<Relation> relations = Utils Dataset<Relation> relations = Utils
.readPath(spark, relationPath, Relation.class) .readPath(spark, graphPath + "/relation", Relation.class)
.filter( .filter(
"dataInfo.deletedbyinference = false and lower(relClass) = '" "dataInfo.deletedbyinference = false and lower(relClass) = '"
+ ModelConstants.IS_PRODUCED_BY.toLowerCase() + "'"); + ModelConstants.IS_PRODUCED_BY.toLowerCase() + "'");
relations spark
.joinWith( .sql(
results, relations.col("source").equalTo(results.col("id")), "Select res.* " +
"inner") "from relation rel " +
.groupByKey( "join result res " +
(MapFunction<Tuple2<Relation, R>, String>) value -> value "on rel.source = res.id " +
._2() "join project p " +
.getId(), "on rel.target = p.id " +
Encoders.STRING()) "")
.mapGroups((MapGroupsFunction<String, Tuple2<Relation, R>, R>) (k, it) -> { .as(Encoders.bean(inputClazz))
return it.next()._2();
}, Encoders.bean(inputClazz))
.write() .write()
.mode(SaveMode.Overwrite) .mode(SaveMode.Overwrite)
.option("compression", "gzip") .option("compression", "gzip")