2021-10-20 17:37:42 +02:00
|
|
|
package eu.dnetlib.dhp.datacite
|
2021-01-28 16:34:46 +01:00
|
|
|
|
2021-10-20 17:12:08 +02:00
|
|
|
import com.fasterxml.jackson.databind.ObjectMapper
|
2021-01-28 16:34:46 +01:00
|
|
|
import eu.dnetlib.dhp.application.ArgumentApplicationParser
|
2021-10-20 17:12:08 +02:00
|
|
|
import eu.dnetlib.dhp.collection.CollectionUtils.fixRelations
|
2021-10-20 17:37:42 +02:00
|
|
|
import eu.dnetlib.dhp.common.Constants.MDSTORE_DATA_PATH
|
|
|
|
import eu.dnetlib.dhp.common.Constants.MDSTORE_SIZE_PATH
|
|
|
|
import eu.dnetlib.dhp.common.vocabulary.VocabularyGroup
|
2021-10-20 17:12:08 +02:00
|
|
|
import eu.dnetlib.dhp.schema.mdstore.{MDStoreVersion, MetadataRecord}
|
2021-01-28 16:34:46 +01:00
|
|
|
import eu.dnetlib.dhp.schema.oaf.Oaf
|
2021-10-20 17:12:08 +02:00
|
|
|
import eu.dnetlib.dhp.utils.DHPUtils.writeHdfsFile
|
2021-01-28 16:34:46 +01:00
|
|
|
import eu.dnetlib.dhp.utils.ISLookupClientFactory
|
|
|
|
import org.apache.spark.SparkConf
|
|
|
|
import org.apache.spark.sql.{Encoder, Encoders, SaveMode, SparkSession}
|
|
|
|
import org.slf4j.{Logger, LoggerFactory}
|
|
|
|
|
|
|
|
import scala.io.Source
|
|
|
|
|
|
|
|
object GenerateDataciteDatasetSpark {
|
|
|
|
|
|
|
|
val log: Logger = LoggerFactory.getLogger(GenerateDataciteDatasetSpark.getClass)
|
|
|
|
|
|
|
|
def main(args: Array[String]): Unit = {
|
|
|
|
val conf = new SparkConf
|
2021-10-20 17:37:42 +02:00
|
|
|
val parser = new ArgumentApplicationParser(Source.fromInputStream(getClass.getResourceAsStream("/eu/dnetlib/dhp/datacite/generate_dataset_params.json")).mkString)
|
2021-01-28 16:34:46 +01:00
|
|
|
parser.parseArgument(args)
|
|
|
|
val master = parser.get("master")
|
|
|
|
val sourcePath = parser.get("sourcePath")
|
2021-06-04 10:14:22 +02:00
|
|
|
val exportLinks = "true".equalsIgnoreCase(parser.get("exportLinks"))
|
2021-01-28 16:34:46 +01:00
|
|
|
val isLookupUrl: String = parser.get("isLookupUrl")
|
|
|
|
log.info("isLookupUrl: {}", isLookupUrl)
|
|
|
|
|
|
|
|
val isLookupService = ISLookupClientFactory.getLookUpService(isLookupUrl)
|
|
|
|
val vocabularies = VocabularyGroup.loadVocsFromIS(isLookupService)
|
|
|
|
val spark: SparkSession = SparkSession.builder().config(conf)
|
|
|
|
.appName(GenerateDataciteDatasetSpark.getClass.getSimpleName)
|
|
|
|
.master(master)
|
|
|
|
.getOrCreate()
|
|
|
|
|
2021-10-20 17:37:42 +02:00
|
|
|
import spark.implicits._
|
|
|
|
|
2021-01-28 16:34:46 +01:00
|
|
|
implicit val mrEncoder: Encoder[MetadataRecord] = Encoders.kryo[MetadataRecord]
|
|
|
|
|
|
|
|
implicit val resEncoder: Encoder[Oaf] = Encoders.kryo[Oaf]
|
|
|
|
|
2021-10-20 17:12:08 +02:00
|
|
|
val mdstoreOutputVersion = parser.get("mdstoreOutputVersion")
|
|
|
|
val mapper = new ObjectMapper()
|
|
|
|
val cleanedMdStoreVersion = mapper.readValue(mdstoreOutputVersion, classOf[MDStoreVersion])
|
|
|
|
val outputBasePath = cleanedMdStoreVersion.getHdfsPath
|
|
|
|
|
|
|
|
log.info("outputBasePath: {}", outputBasePath)
|
|
|
|
val targetPath = s"$outputBasePath/$MDSTORE_DATA_PATH"
|
|
|
|
|
2021-01-28 16:34:46 +01:00
|
|
|
spark.read.load(sourcePath).as[DataciteType]
|
|
|
|
.filter(d => d.isActive)
|
2021-06-04 10:14:22 +02:00
|
|
|
.flatMap(d => DataciteToOAFTransformation.generateOAF(d.json, d.timestamp, d.timestamp, vocabularies, exportLinks))
|
2021-01-28 16:34:46 +01:00
|
|
|
.filter(d => d != null)
|
2021-10-20 17:37:42 +02:00
|
|
|
.flatMap(i => fixRelations(i)).filter(i => i != null)
|
2021-01-28 16:34:46 +01:00
|
|
|
.write.mode(SaveMode.Overwrite).save(targetPath)
|
2021-10-20 17:12:08 +02:00
|
|
|
|
2021-10-20 17:37:42 +02:00
|
|
|
val total_items = spark.read.load(targetPath).as[Oaf].count()
|
2021-10-20 17:12:08 +02:00
|
|
|
writeHdfsFile(spark.sparkContext.hadoopConfiguration, s"$total_items", outputBasePath + MDSTORE_SIZE_PATH)
|
2021-01-28 16:34:46 +01:00
|
|
|
}
|
2021-10-20 17:37:42 +02:00
|
|
|
}
|