80 lines
2.6 KiB
Scala
80 lines
2.6 KiB
Scala
package eu.dnetlib.dhp.sx.graph
|
|
|
|
import eu.dnetlib.dhp.application.ArgumentApplicationParser
|
|
import eu.dnetlib.dhp.schema.oaf.{Oaf, OtherResearchProduct, Publication, Relation, Result, Software, Dataset => OafDataset}
|
|
import org.apache.commons.io.IOUtils
|
|
import org.apache.spark.SparkConf
|
|
import org.apache.spark.sql.{Dataset, Encoder, Encoders, SaveMode, SparkSession}
|
|
import org.slf4j.{Logger, LoggerFactory}
|
|
|
|
|
|
|
|
object SparkCreateInputGraph {
|
|
|
|
def main(args: Array[String]): Unit = {
|
|
|
|
val log: Logger = LoggerFactory.getLogger(getClass)
|
|
val conf: SparkConf = new SparkConf()
|
|
val parser = new ArgumentApplicationParser(IOUtils.toString(getClass.getResourceAsStream("/eu/dnetlib/dhp/sx/graph/extract_entities_params.json")))
|
|
parser.parseArgument(args)
|
|
val spark: SparkSession =
|
|
SparkSession
|
|
.builder()
|
|
.config(conf)
|
|
.appName(getClass.getSimpleName)
|
|
.master(parser.get("master")).getOrCreate()
|
|
|
|
|
|
val resultObject = List(
|
|
("publication", classOf[Publication]),
|
|
("dataset", classOf[OafDataset]),
|
|
("software", classOf[Software]),
|
|
("otherResearchProduct", classOf[OtherResearchProduct])
|
|
|
|
)
|
|
|
|
implicit val oafEncoder: Encoder[Oaf] = Encoders.kryo(classOf[Oaf])
|
|
|
|
val sourcePath = parser.get("sourcePath")
|
|
log.info(s"sourcePath -> $sourcePath")
|
|
val targetPath = parser.get("targetPath")
|
|
log.info(s"targetPath -> $targetPath")
|
|
|
|
|
|
val oafDs:Dataset[Oaf] = spark.read.load(s"$sourcePath/*").as[Oaf]
|
|
|
|
resultObject.foreach(r => extractEntities(oafDs,s"$targetPath/extracted/${r._1}", r._2, log))
|
|
|
|
extractEntities(oafDs,s"$targetPath/extracted/relation", classOf[Relation], log)
|
|
|
|
resultObject.foreach { r =>
|
|
log.info(s"Make ${r._1} unique")
|
|
makeDatasetUnique(s"$targetPath/extracted/${r._1}",s"$targetPath/dedup/${r._1}",spark, r._2)
|
|
}
|
|
}
|
|
|
|
|
|
def extractEntities[T <: Oaf ](oafDs:Dataset[Oaf], targetPath:String, clazz:Class[T], log:Logger) :Unit = {
|
|
|
|
implicit val resEncoder: Encoder[T] = Encoders.kryo(clazz)
|
|
log.info(s"Extract ${clazz.getSimpleName}")
|
|
oafDs.filter(o => o.isInstanceOf[T]).map(p => p.asInstanceOf[T]).write.mode(SaveMode.Overwrite).save(targetPath)
|
|
}
|
|
|
|
|
|
def makeDatasetUnique[T <: Result ](sourcePath:String, targetPath:String, spark:SparkSession, clazz:Class[T]) :Unit = {
|
|
import spark.implicits._
|
|
|
|
implicit val resEncoder: Encoder[T] = Encoders.kryo(clazz)
|
|
|
|
val ds:Dataset[T] = spark.read.load(sourcePath).as[T]
|
|
|
|
ds.groupByKey(_.getId).reduceGroups{(x,y) =>
|
|
x.mergeFrom(y)
|
|
x
|
|
}.map(_._2).write.mode(SaveMode.Overwrite).save(targetPath)
|
|
|
|
}
|
|
|
|
}
|