dnet-hadoop/dhp-workflows/dhp-doiboost/src/main/java/eu/dnetlib/doiboost/crossref/SparkMapDumpIntoOAF.scala

package eu.dnetlib.doiboost.crossref

import eu.dnetlib.dhp.application.ArgumentApplicationParser
import eu.dnetlib.dhp.schema.oaf
import eu.dnetlib.dhp.schema.oaf.{Oaf, Publication, Relation, Dataset => OafDataset}
import org.apache.commons.io.IOUtils
import org.apache.hadoop.io.{IntWritable, Text}
import org.apache.spark.SparkConf
import org.apache.spark.rdd.RDD
import org.apache.spark.sql.{Dataset, Encoder, Encoders, SaveMode, SparkSession}
import org.slf4j.{Logger, LoggerFactory}


case class Reference(author: String, firstPage: String) {}

object SparkMapDumpIntoOAF {

  def main(args: Array[String]): Unit = {


    val logger: Logger = LoggerFactory.getLogger(SparkMapDumpIntoOAF.getClass)
    val conf: SparkConf = new SparkConf()
    val parser = new ArgumentApplicationParser(IOUtils.toString(SparkMapDumpIntoOAF.getClass.getResourceAsStream("/eu/dnetlib/dhp/doiboost/convert_map_to_oaf_params.json")))
    parser.parseArgument(args)
    val spark: SparkSession =
      SparkSession
        .builder()
        .config(conf)
        .appName(SparkMapDumpIntoOAF.getClass.getSimpleName)
        .master(parser.get("master")).getOrCreate()

    implicit val mapEncoderPubs: Encoder[Publication] = Encoders.kryo[Publication]
    implicit val mapEncoderRelatons: Encoder[Relation] = Encoders.kryo[Relation]
    implicit val mapEncoderDatasets: Encoder[oaf.Dataset] = Encoders.kryo[OafDataset]

    val sc = spark.sparkContext
    val targetPath = parser.get("targetPath")


    sc.sequenceFile(parser.get("sourcePath"), classOf[IntWritable], classOf[Text])
      .map(k => k._2.toString).map(CrossrefImporter.decompressBlob)
      .flatMap(k => Crossref2Oaf.convert(k)).saveAsObjectFile(s"${targetPath}/mixObject")

    val inputRDD = sc.objectFile[Oaf](s"${targetPath}/mixObject").filter(p=> p!= null)

    val distinctPubs:RDD[Publication] = inputRDD.filter(k => k != null && k.isInstanceOf[Publication])
      .map(k => k.asInstanceOf[Publication]).map { p: Publication => Tuple2(p.getId, p) }.reduceByKey { case (p1: Publication, p2: Publication) =>
      var r = if (p1 == null) p2 else p1
      if (p1 != null && p2 != null) {
        if (p1.getLastupdatetimestamp != null && p2.getLastupdatetimestamp != null) {
          if (p1.getLastupdatetimestamp < p2.getLastupdatetimestamp)
            r = p2
          else
            r = p1
        } else {
          r = if (p1.getLastupdatetimestamp == null) p2 else p1
        }
      }
      r
    }.map(_._2)

    val pubs:Dataset[Publication] = spark.createDataset(distinctPubs)
    pubs.write.mode(SaveMode.Overwrite).save(s"${targetPath}/publication")


    val distincDatasets:RDD[OafDataset] = inputRDD.filter(k => k != null && k.isInstanceOf[OafDataset])
      .map(k => k.asInstanceOf[OafDataset]).map(p => Tuple2(p.getId, p)).reduceByKey { case (p1: OafDataset, p2: OafDataset) =>
      var r = if (p1 == null) p2 else p1
      if (p1 != null && p2 != null) {
        if (p1.getLastupdatetimestamp != null && p2.getLastupdatetimestamp != null) {
          if (p1.getLastupdatetimestamp < p2.getLastupdatetimestamp)
            r = p2
          else
            r = p1
        } else {
          r = if (p1.getLastupdatetimestamp == null) p2 else p1
        }
      }
      r
    }.map(_._2)

    spark.createDataset(distincDatasets).write.mode(SaveMode.Overwrite).save(s"${targetPath}/dataset")



    val distinctRels =inputRDD.filter(k => k != null && k.isInstanceOf[Relation])
      .map(k => k.asInstanceOf[Relation]).map(r=> (s"${r.getSource}::${r.getTarget}",r))
      .reduceByKey { case (p1: Relation, p2: Relation) =>
        if (p1 == null) p2 else p1
      }.map(_._2)

    val rels: Dataset[Relation] = spark.createDataset(distinctRels)

    rels.write.mode(SaveMode.Overwrite).save(s"${targetPath}/relations")
  }


}