added a sequentialization step on the spark job. Addedd new parameter

This commit is contained in:
Miriam Baglioni 2020-05-05 17:03:43 +02:00
parent 252b219dd5
commit dd2e698a72
4 changed files with 133 additions and 47 deletions

View File

@ -0,0 +1,4 @@
package eu.dnetlib.dhp.countrypropagation;
public class PrepareResultCountrySet {
}

View File

@ -3,8 +3,10 @@ package eu.dnetlib.dhp.countrypropagation;
import static eu.dnetlib.dhp.PropagationConstant.*;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkHiveSession;
import static jdk.nashorn.internal.objects.NativeDebug.map;
import java.util.*;
import java.util.stream.Collectors;
import org.apache.commons.io.IOUtils;
import org.apache.spark.SparkConf;
@ -52,6 +54,11 @@ public class SparkCountryPropagationJob2 {
final String datasourcecountrypath = parser.get("preparedInfoPath");
log.info("preparedInfoPath: {}", datasourcecountrypath);
final String possibleUpdatesPath = datasourcecountrypath
.substring(0, datasourcecountrypath.lastIndexOf("/") + 1)
+ "possibleUpdates";
log.info("possibleUpdatesPath: {}", possibleUpdatesPath);
final String resultClassName = parser.get("resultTableName");
log.info("resultTableName: {}", resultClassName);
@ -70,13 +77,14 @@ public class SparkCountryPropagationJob2 {
conf,
isSparkSessionManaged,
spark -> {
removeOutputDir(spark, possibleUpdatesPath);
execPropagation(
spark,
datasourcecountrypath,
inputPath,
outputPath,
resultClazz,
saveGraph);
saveGraph, possibleUpdatesPath);
});
}
@ -86,19 +94,30 @@ public class SparkCountryPropagationJob2 {
String inputPath,
String outputPath,
Class<R> resultClazz,
boolean saveGraph) {
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
boolean saveGraph, String possilbeUpdatesPath) {
// final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
// Load file with preprocessed association datasource - country
Dataset<DatasourceCountry> datasourcecountryassoc = readAssocDatasourceCountry(spark, datasourcecountrypath);
// broadcasting the result of the preparation step
Broadcast<Dataset<DatasourceCountry>> broadcast_datasourcecountryassoc = sc.broadcast(datasourcecountryassoc);
// Broadcast<Dataset<DatasourceCountry>> broadcast_datasourcecountryassoc =
// sc.broadcast(datasourcecountryassoc);
Dataset<ResultCountrySet> potentialUpdates = getPotentialResultToUpdate(
spark, inputPath, resultClazz, broadcast_datasourcecountryassoc)
spark, inputPath, resultClazz, datasourcecountryassoc)
.as(Encoders.bean(ResultCountrySet.class));
potentialUpdates.write().option("compression", "gzip").mode(SaveMode.Overwrite).json(possilbeUpdatesPath);
if (saveGraph) {
// updateResultTable(spark, potentialUpdates, inputPath, resultClazz, outputPath);
potentialUpdates = spark
.read()
.textFile(possilbeUpdatesPath)
.map(
(MapFunction<String, ResultCountrySet>) value -> OBJECT_MAPPER
.readValue(value, ResultCountrySet.class),
Encoders.bean(ResultCountrySet.class));
updateResultTable(spark, potentialUpdates, inputPath, resultClazz, outputPath);
}
}
@ -113,69 +132,116 @@ public class SparkCountryPropagationJob2 {
log.info("Reading Graph table from: {}", inputPath);
Dataset<R> result = readPathEntity(spark, inputPath, resultClazz);
Dataset<Tuple2<String, R>> result_pair = result
.map(
r -> new Tuple2<>(r.getId(), r),
Encoders.tuple(Encoders.STRING(), Encoders.bean(resultClazz)));
Dataset<R> new_table = result_pair
Dataset<R> new_table = result
.joinWith(
potentialUpdates,
result_pair.col("_1").equalTo(potentialUpdates.col("resultId")),
potentialUpdates, result
.col("id")
.equalTo(potentialUpdates.col("resultId")),
"left_outer")
.map(
(MapFunction<Tuple2<Tuple2<String, R>, ResultCountrySet>, R>) value -> {
R r = value._1()._2();
.map((MapFunction<Tuple2<R, ResultCountrySet>, R>) value -> {
R r = value._1();
Optional<ResultCountrySet> potentialNewCountries = Optional.ofNullable(value._2());
if (potentialNewCountries.isPresent()) {
HashSet<String> countries = new HashSet<>();
for (Qualifier country : r.getCountry()) {
countries.add(country.getClassid());
}
Result res = new Result();
res.setId(r.getId());
List<Country> countryList = new ArrayList<>();
for (CountrySbs country : potentialNewCountries
HashSet<String> countries = r
.getCountry()
.stream()
.map(c -> c.getClassid())
.collect(Collectors.toCollection(HashSet::new));
r
.getCountry()
.addAll(
potentialNewCountries
.get()
.getCountrySet()) {
if (!countries.contains(country.getClassid())) {
countryList
.add(
getCountry(
country.getClassid(),
country.getClassname()));
}
}
res.setCountry(countryList);
r.mergeFrom(res);
.getCountrySet()
.stream()
.filter(c -> !countries.contains(c.getClassid()))
.map(c -> getCountry(c.getClassid(), c.getClassname()))
.collect(Collectors.toList()));
// Result res = new Result();
// res.setId(r.getId());
// List<Country> countryList = new ArrayList<>();
// for (CountrySbs country : potentialNewCountries
// .get()
// .getCountrySet()) {
// if (!countries.contains(country.getClassid())) {
// countryList
// .add(
// getCountry(
// country.getClassid(),
// country.getClassname()));
// }
// }
// res.setCountry(countryList);
// r.mergeFrom(res);
}
return r;
},
Encoders.bean(resultClazz));
}, Encoders.bean(resultClazz));
// Dataset<Tuple2<String, R>> result_pair = result
// .map(
// r -> new Tuple2<>(r.getId(), r),
// Encoders.tuple(Encoders.STRING(), Encoders.bean(resultClazz)));
//
// Dataset<R> new_table = result_pair
// .joinWith(
// potentialUpdates,
// result_pair.col("_1").equalTo(potentialUpdates.col("resultId")),
// "left_outer")
// .map(
// (MapFunction<Tuple2<Tuple2<String, R>, ResultCountrySet>, R>) value -> {
// R r = value._1()._2();
// Optional<ResultCountrySet> potentialNewCountries = Optional.ofNullable(value._2());
// if (potentialNewCountries.isPresent()) {
// HashSet<String> countries = new HashSet<>();
// for (Qualifier country : r.getCountry()) {
// countries.add(country.getClassid());
// }
// Result res = new Result();
// res.setId(r.getId());
// List<Country> countryList = new ArrayList<>();
// for (CountrySbs country : potentialNewCountries
// .get()
// .getCountrySet()) {
// if (!countries.contains(country.getClassid())) {
// countryList
// .add(
// getCountry(
// country.getClassid(),
// country.getClassname()));
// }
// }
// res.setCountry(countryList);
// r.mergeFrom(res);
// }
// return r;
// },
// Encoders.bean(resultClazz));
log.info("Saving graph table to path: {}", outputPath);
// log.info("number of saved recordsa: {}", new_table.count());
new_table.toJSON().write().option("compression", "gzip").text(outputPath);
new_table.write().option("compression", "gzip").mode(SaveMode.Overwrite).json(outputPath);
}
private static <R extends Result> Dataset<Row> getPotentialResultToUpdate(
SparkSession spark,
String inputPath,
Class<R> resultClazz,
Broadcast<Dataset<DatasourceCountry>> broadcast_datasourcecountryassoc) {
Dataset<DatasourceCountry> datasourcecountryassoc) {
Dataset<R> result = readPathEntity(spark, inputPath, resultClazz);
result.createOrReplaceTempView("result");
// log.info("number of results: {}", result.count());
createCfHbforresult(spark);
return countryPropagationAssoc(spark, broadcast_datasourcecountryassoc);
return countryPropagationAssoc(spark, datasourcecountryassoc);
}
private static Dataset<Row> countryPropagationAssoc(
SparkSession spark,
Broadcast<Dataset<DatasourceCountry>> broadcast_datasourcecountryassoc) {
Dataset<DatasourceCountry> datasource_country) {
Dataset<DatasourceCountry> datasource_country = broadcast_datasourcecountryassoc.value();
// Dataset<DatasourceCountry> datasource_country = broadcast_datasourcecountryassoc.value();
datasource_country.createOrReplaceTempView("datasource_country");
log.info("datasource_country number : {}", datasource_country.count());

View File

@ -140,6 +140,7 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.dynamicAllocation.enabled=true
--conf spark.dynamicAllocation.maxExecutors=${spark2MaxExecutors}
--conf spark.speculation=false
</spark-opts>
<arg>--sourcePath</arg><arg>${sourcePath}/publication</arg>
<arg>--hive_metastore_uris</arg><arg>${hive_metastore_uris}</arg>
@ -169,6 +170,7 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.dynamicAllocation.enabled=true
--conf spark.dynamicAllocation.maxExecutors=${spark2MaxExecutors}
--conf spark.speculation=false
</spark-opts>
<arg>--sourcePath</arg><arg>${sourcePath}/dataset</arg>
<arg>--hive_metastore_uris</arg><arg>${hive_metastore_uris}</arg>
@ -198,6 +200,7 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.dynamicAllocation.enabled=true
--conf spark.dynamicAllocation.maxExecutors=${spark2MaxExecutors}
--conf spark.speculation=false
</spark-opts>
<arg>--sourcePath</arg><arg>${sourcePath}/otherresearchproduct</arg>
<arg>--hive_metastore_uris</arg><arg>${hive_metastore_uris}</arg>
@ -227,6 +230,7 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.dynamicAllocation.enabled=true
--conf spark.dynamicAllocation.maxExecutors=${spark2MaxExecutors}
--conf spark.speculation=false
</spark-opts>
<arg>--sourcePath</arg><arg>${sourcePath}/software</arg>
<arg>--hive_metastore_uris</arg><arg>${hive_metastore_uris}</arg>

View File

@ -261,6 +261,9 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.dynamicAllocation.enabled=true
--conf spark.dynamicAllocation.maxExecutors=${spark2MaxExecutors}
--conf spark.speculation=false
--conf spark.hadoop.mapreduce.map.speculative=false
--conf spark.hadoop.mapreduce.reduce.speculative=false
</spark-opts>
<arg>--possibleUpdatesPath</arg><arg>${workingDir}/preparedInfo/mergedOrcidAssoc</arg>
<arg>--sourcePath</arg><arg>${sourcePath}/publication</arg>
@ -289,6 +292,9 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.dynamicAllocation.enabled=true
--conf spark.dynamicAllocation.maxExecutors=${spark2MaxExecutors}
--conf spark.speculation=false
--conf spark.hadoop.mapreduce.map.speculative=false
--conf spark.hadoop.mapreduce.reduce.speculative=false
</spark-opts>
<arg>--possibleUpdatesPath</arg><arg>${workingDir}/preparedInfo/mergedOrcidAssoc</arg>
<arg>--sourcePath</arg><arg>${sourcePath}/dataset</arg>
@ -317,6 +323,9 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.dynamicAllocation.enabled=true
--conf spark.dynamicAllocation.maxExecutors=${spark2MaxExecutors}
--conf spark.speculation=false
--conf spark.hadoop.mapreduce.map.speculative=false
--conf spark.hadoop.mapreduce.reduce.speculative=false
</spark-opts>
<arg>--possibleUpdatesPath</arg><arg>${workingDir}/preparedInfo/mergedOrcidAssoc</arg>
<arg>--sourcePath</arg><arg>${sourcePath}/otherresearchproduct</arg>
@ -345,6 +354,9 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.dynamicAllocation.enabled=true
--conf spark.dynamicAllocation.maxExecutors=${spark2MaxExecutors}
--conf spark.speculation=false
--conf spark.hadoop.mapreduce.map.speculative=false
--conf spark.hadoop.mapreduce.reduce.speculative=false
</spark-opts>
<arg>--possibleUpdatesPath</arg><arg>${workingDir}/preparedInfo/mergedOrcidAssoc</arg>
<arg>--sourcePath</arg><arg>${sourcePath}/software</arg>