This commit is contained in:
Claudio Atzori 2020-12-07 13:50:08 +01:00
commit 5e8509bef7
8 changed files with 107 additions and 144 deletions

View File

@ -6,7 +6,7 @@
<groupId>eu.dnetlib.dhp</groupId> <groupId>eu.dnetlib.dhp</groupId>
<artifactId>dhp</artifactId> <artifactId>dhp</artifactId>
<version>1.2.4-SNAPSHOT</version> <version>1.2.4-SNAPSHOT</version>
<relativePath>../</relativePath> <relativePath>../pom.xml</relativePath>
</parent> </parent>
<artifactId>dhp-schemas</artifactId> <artifactId>dhp-schemas</artifactId>

View File

@ -0,0 +1,24 @@
package eu.dnetlib.dhp.schema.orcid;
import java.util.List;
public class OrcidDOI {
private String doi;
private List<AuthorData> authors;
public String getDoi() {
return doi;
}
public void setDoi(String doi) {
this.doi = doi;
}
public List<AuthorData> getAuthors() {
return authors;
}
public void setAuthors(List<AuthorData> authors) {
this.authors = authors;
}
}

View File

@ -200,7 +200,7 @@ case object Crossref2Oaf {
a.setSurname(family) a.setSurname(family)
a.setFullname(s"$given $family") a.setFullname(s"$given $family")
if (StringUtils.isNotBlank(orcid)) if (StringUtils.isNotBlank(orcid))
a.setPid(List(createSP(orcid, ORCID, PID_TYPES)).asJava) a.setPid(List(createSP(orcid, ORCID, PID_TYPES, generateDataInfo())).asJava)
a a
} }
@ -248,7 +248,7 @@ case object Crossref2Oaf {
def snsfRule(award:String): String = { def snsfRule(award:String): String = {
var tmp1 = StringUtils.substringAfter(award,"_") val tmp1 = StringUtils.substringAfter(award,"_")
val tmp2 = StringUtils.substringBefore(tmp1,"/") val tmp2 = StringUtils.substringBefore(tmp1,"/")
logger.debug(s"From $award to $tmp2") logger.debug(s"From $award to $tmp2")
tmp2 tmp2
@ -294,7 +294,7 @@ case object Crossref2Oaf {
} }
def getProjectId (nsPrefix:String, targetId:String):String = { def getProjectId (nsPrefix:String, targetId:String):String = {
"40|$nsPrefix::$targetId" s"40|$nsPrefix::$targetId"
} }

View File

@ -2,6 +2,7 @@ package eu.dnetlib.doiboost.crossref
import eu.dnetlib.dhp.application.ArgumentApplicationParser import eu.dnetlib.dhp.application.ArgumentApplicationParser
import org.apache.commons.io.IOUtils import org.apache.commons.io.IOUtils
import org.apache.hadoop.io.{IntWritable, Text}
import org.apache.spark.SparkConf import org.apache.spark.SparkConf
import org.apache.spark.sql.expressions.Aggregator import org.apache.spark.sql.expressions.Aggregator
import org.apache.spark.sql.{Dataset, Encoder, Encoders, SaveMode, SparkSession} import org.apache.spark.sql.{Dataset, Encoder, Encoders, SaveMode, SparkSession}
@ -12,21 +13,23 @@ import org.slf4j.{Logger, LoggerFactory}
object CrossrefDataset { object CrossrefDataset {
val logger: Logger = LoggerFactory.getLogger(SparkMapDumpIntoOAF.getClass)
def extractTimestamp(input:String): Long = {
def to_item(input:String):CrossrefDT = {
implicit lazy val formats: DefaultFormats.type = org.json4s.DefaultFormats implicit lazy val formats: DefaultFormats.type = org.json4s.DefaultFormats
lazy val json: json4s.JValue = parse(input) lazy val json: json4s.JValue = parse(input)
val ts:Long = (json \ "indexed" \ "timestamp").extract[Long]
(json\"indexed"\"timestamp").extractOrElse[Long](0) val doi:String = (json \ "DOI").extract[String]
CrossrefDT(doi, input, ts)
} }
def main(args: Array[String]): Unit = { def main(args: Array[String]): Unit = {
val logger: Logger = LoggerFactory.getLogger(SparkMapDumpIntoOAF.getClass)
val conf: SparkConf = new SparkConf() val conf: SparkConf = new SparkConf()
val parser = new ArgumentApplicationParser(IOUtils.toString(CrossrefDataset.getClass.getResourceAsStream("/eu/dnetlib/dhp/doiboost/crossref_to_dataset_params.json"))) val parser = new ArgumentApplicationParser(IOUtils.toString(CrossrefDataset.getClass.getResourceAsStream("/eu/dnetlib/dhp/doiboost/crossref_to_dataset_params.json")))
parser.parseArgument(args) parser.parseArgument(args)
@ -49,9 +52,8 @@ object CrossrefDataset {
if (a == null) if (a == null)
return b return b
val tb = extractTimestamp(b.json)
val ta = extractTimestamp(a.json) if(a.timestamp >b.timestamp) {
if(ta >tb) {
return a return a
} }
b b
@ -63,9 +65,7 @@ object CrossrefDataset {
if (a == null) if (a == null)
return b return b
val tb = extractTimestamp(b.json) if(a.timestamp >b.timestamp) {
val ta = extractTimestamp(a.json)
if(ta >tb) {
return a return a
} }
b b
@ -78,15 +78,21 @@ object CrossrefDataset {
override def finish(reduction: CrossrefDT): CrossrefDT = reduction override def finish(reduction: CrossrefDT): CrossrefDT = reduction
} }
val sourcePath:String = parser.get("sourcePath") val workingPath:String = parser.get("workingPath")
val targetPath:String = parser.get("targetPath")
val ds:Dataset[CrossrefDT] = spark.read.load(sourcePath).as[CrossrefDT]
ds.groupByKey(_.doi) val main_ds:Dataset[CrossrefDT] = spark.read.load(s"$workingPath/crossref_ds").as[CrossrefDT]
val update =
spark.createDataset(spark.sparkContext.sequenceFile(s"$workingPath/index_update", classOf[IntWritable], classOf[Text])
.map(i =>CrossrefImporter.decompressBlob(i._2.toString))
.map(i =>to_item(i)))
main_ds.union(update).groupByKey(_.doi)
.agg(crossrefAggregator.toColumn) .agg(crossrefAggregator.toColumn)
.map(s=>s._2) .map(s=>s._2)
.write.mode(SaveMode.Overwrite).save(targetPath) .write.mode(SaveMode.Overwrite).save(s"$workingPath/crossref_ds_updated")
} }

View File

@ -34,85 +34,21 @@ object SparkMapDumpIntoOAF {
implicit val mapEncoderRelatons: Encoder[Relation] = Encoders.kryo[Relation] implicit val mapEncoderRelatons: Encoder[Relation] = Encoders.kryo[Relation]
implicit val mapEncoderDatasets: Encoder[oaf.Dataset] = Encoders.kryo[OafDataset] implicit val mapEncoderDatasets: Encoder[oaf.Dataset] = Encoders.kryo[OafDataset]
val sc = spark.sparkContext
val targetPath = parser.get("targetPath") val targetPath = parser.get("targetPath")
import spark.implicits._ import spark.implicits._
spark.read.load(parser.get("sourcePath")).as[CrossrefDT] spark.read.load(parser.get("sourcePath")).as[CrossrefDT]
.flatMap(k => Crossref2Oaf.convert(k.json)) .flatMap(k => Crossref2Oaf.convert(k.json))
.filter(o => o != null) .filter(o => o != null)
.write.mode(SaveMode.Overwrite).save(s"$targetPath/mixObject") .write.mode(SaveMode.Overwrite).save(s"$targetPath/mixObject")
val ds:Dataset[Oaf] = spark.read.load(s"$targetPath/mixObject").as[Oaf] val ds:Dataset[Oaf] = spark.read.load(s"$targetPath/mixObject").as[Oaf]
ds.filter(o => o.isInstanceOf[Publication]).map(o => o.asInstanceOf[Publication]).write.save(s"$targetPath/publication") ds.filter(o => o.isInstanceOf[Publication]).map(o => o.asInstanceOf[Publication]).write.mode(SaveMode.Overwrite).save(s"$targetPath/crossrefPublication")
ds.filter(o => o.isInstanceOf[Relation]).map(o => o.asInstanceOf[Relation]).write.save(s"$targetPath/relation") ds.filter(o => o.isInstanceOf[Relation]).map(o => o.asInstanceOf[Relation]).write.mode(SaveMode.Overwrite).save(s"$targetPath/crossrefRelation")
ds.filter(o => o.isInstanceOf[OafDataset]).map(o => o.asInstanceOf[OafDataset]).write.save(s"$targetPath/dataset") ds.filter(o => o.isInstanceOf[OafDataset]).map(o => o.asInstanceOf[OafDataset]).write.mode(SaveMode.Overwrite).save(s"$targetPath/crossrefDataset")
//
//
//
// 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")
} }

View File

@ -16,88 +16,86 @@
<name>sparkExecutorCores</name> <name>sparkExecutorCores</name>
<description>number of cores used by single executor</description> <description>number of cores used by single executor</description>
</property> </property>
<!-- <property>--> <property>
<!-- <name>timestamp</name>--> <name>timestamp</name>
<!-- <description>Timestamp for incremental Harvesting</description>--> <description>Timestamp for incremental Harvesting</description>
<!-- </property>--> </property>
</parameters> </parameters>
<start to="ExtractCrossrefToOAF"/> <start to="ImportCrossRef"/>
<kill name="Kill"> <kill name="Kill">
<message>Action failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message> <message>Action failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
</kill> </kill>
<!-- <action name="ResetWorkingPath">--> <action name="ImportCrossRef">
<!-- <fs>--> <java>
<!-- <delete path='${workingPath}/input/crossref/index_dump'/>--> <job-tracker>${jobTracker}</job-tracker>
<!--&lt;!&ndash; <mkdir path='${workingPath}/input/crossref'/>&ndash;&gt;--> <name-node>${nameNode}</name-node>
<!-- </fs>--> <main-class>eu.dnetlib.doiboost.crossref.CrossrefImporter</main-class>
<!-- <ok to="ImportCrossRef"/>--> <arg>-t</arg><arg>${workingPath}/input/crossref/index_update</arg>
<!-- <error to="Kill"/>--> <arg>-n</arg><arg>${nameNode}</arg>
<!-- </action>--> <arg>-ts</arg><arg>${timestamp}</arg>
</java>
<ok to="GenerateDataset"/>
<error to="Kill"/>
</action>
<action name="GenerateDataset">
<!-- <action name="ImportCrossRef">-->
<!-- <java>-->
<!-- <job-tracker>${jobTracker}</job-tracker>-->
<!-- <name-node>${nameNode}</name-node>-->
<!-- <main-class>eu.dnetlib.doiboost.crossref.CrossrefImporter</main-class>-->
<!-- <arg>-t</arg><arg>${workingPath}/input/crossref/index_dump_1</arg>-->
<!-- <arg>-n</arg><arg>${nameNode}</arg>-->
<!-- <arg>-ts</arg><arg>${timestamp}</arg>-->
<!-- </java>-->
<!-- <ok to="End"/>-->
<!-- <error to="Kill"/>-->
<!-- </action>-->
<action name="ExtractCrossrefToOAF">
<spark xmlns="uri:oozie:spark-action:0.2"> <spark xmlns="uri:oozie:spark-action:0.2">
<master>yarn-cluster</master> <master>yarn-cluster</master>
<mode>cluster</mode> <mode>cluster</mode>
<name>ExtractCrossrefToOAF</name> <name>ExtractCrossrefToOAF</name>
<class>eu.dnetlib.doiboost.crossref.CrossrefDataset</class>
<jar>dhp-doiboost-${projectVersion}.jar</jar>
<spark-opts>
--executor-memory=${sparkExecutorMemory}
--executor-cores=${sparkExecutorCores}
--driver-memory=${sparkDriverMemory}
--conf spark.sql.shuffle.partitions=3840
${sparkExtraOPT}
</spark-opts>
<arg>--workingPath</arg><arg>/data/doiboost/input/crossref</arg>
<arg>--master</arg><arg>yarn-cluster</arg>
</spark>
<ok to="RenameDataset"/>
<error to="Kill"/>
</action>
<action name="RenameDataset">
<fs>
<delete path='${workingPath}/input/crossref/crossref_ds'/>
<move source="${workingPath}/input/crossref/crossref_ds_updated"
target="${workingPath}/input/crossref/crossref_ds"/>
</fs>
<ok to="ConvertCrossrefToOAF"/>
<error to="Kill"/>
</action>
<action name="ConvertCrossrefToOAF">
<spark xmlns="uri:oozie:spark-action:0.2">
<master>yarn-cluster</master>
<mode>cluster</mode>
<name>ConvertCrossrefToOAF</name>
<class>eu.dnetlib.doiboost.crossref.SparkMapDumpIntoOAF</class> <class>eu.dnetlib.doiboost.crossref.SparkMapDumpIntoOAF</class>
<jar>dhp-doiboost-${projectVersion}.jar</jar> <jar>dhp-doiboost-${projectVersion}.jar</jar>
<spark-opts> <spark-opts>
--executor-memory=${sparkExecutorMemory} --executor-memory=${sparkExecutorMemory}
--executor-cores=${sparkExecutorCores} --executor-cores=${sparkExecutorCores}
--driver-memory=${sparkDriverMemory} --driver-memory=${sparkDriverMemory}
--conf spark.sql.shuffle.partitions=3840
${sparkExtraOPT} ${sparkExtraOPT}
</spark-opts> </spark-opts>
<arg>--sourcePath</arg><arg>${workingPath}/input/crossref/crossref_ds</arg> <arg>--sourcePath</arg><arg>${workingPath}/input/crossref/crossref_ds</arg>
<arg>--targetPath</arg><arg>${workingPath}/input/crossref</arg> <arg>--targetPath</arg><arg>${workingPath}/process/</arg>
<arg>--master</arg><arg>yarn-cluster</arg> <arg>--master</arg><arg>yarn-cluster</arg>
</spark> </spark>
<ok to="End"/> <ok to="End"/>
<error to="Kill"/> <error to="Kill"/>
</action> </action>
<!-- <action name="GenerateDataset">-->
<!-- <spark xmlns="uri:oozie:spark-action:0.2">-->
<!-- <master>yarn-cluster</master>-->
<!-- <mode>cluster</mode>-->
<!-- <name>ExtractCrossrefToOAF</name>-->
<!-- <class>eu.dnetlib.doiboost.crossref.CrossrefDataset</class>-->
<!-- <jar>dhp-doiboost-${projectVersion}.jar</jar>-->
<!-- <spark-opts>-->
<!-- &#45;&#45;executor-memory=${sparkExecutorMemory}-->
<!-- &#45;&#45;executor-cores=${sparkExecutorCores}-->
<!-- &#45;&#45;driver-memory=${sparkDriverMemory}-->
<!-- ${sparkExtraOPT}-->
<!-- </spark-opts>-->
<!-- <arg>&#45;&#45;sourcePath</arg><arg>/data/doiboost/crossref/cr_dataset</arg>-->
<!-- <arg>&#45;&#45;targetPath</arg><arg>/data/doiboost/crossref/crossrefDataset</arg>-->
<!-- <arg>&#45;&#45;master</arg><arg>yarn-cluster</arg>-->
<!-- </spark>-->
<!-- <ok to="End"/>-->
<!-- <error to="Kill"/>-->
<!-- </action>-->
<end name="End"/> <end name="End"/>
</workflow-app> </workflow-app>

View File

@ -1,6 +1,5 @@
[ [
{"paramName":"s", "paramLongName":"sourcePath", "paramDescription": "the path of the sequencial file to read", "paramRequired": true}, {"paramName":"w", "paramLongName":"workingPath", "paramDescription": "the working dir path", "paramRequired": true},
{"paramName":"t", "paramLongName":"targetPath", "paramDescription": "the working dir path", "paramRequired": true},
{"paramName":"m", "paramLongName":"master", "paramDescription": "the master name", "paramRequired": true} {"paramName":"m", "paramLongName":"master", "paramDescription": "the master name", "paramRequired": true}
] ]

View File

@ -7,7 +7,7 @@
<groupId>eu.dnetlib.dhp</groupId> <groupId>eu.dnetlib.dhp</groupId>
<artifactId>dhp</artifactId> <artifactId>dhp</artifactId>
<version>1.2.4-SNAPSHOT</version> <version>1.2.4-SNAPSHOT</version>
<relativePath>../</relativePath> <relativePath>../pom.xml</relativePath>
</parent> </parent>
<artifactId>dhp-workflows</artifactId> <artifactId>dhp-workflows</artifactId>