new parametric implementation of country propagation. Exploits information compute before and broadcasts it to each executor

This commit is contained in:
Miriam Baglioni 2020-04-11 16:36:59 +02:00
parent a2d833d5dd
commit aef9b3aa90
1 changed files with 70 additions and 36 deletions

View File

@ -4,6 +4,9 @@ import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.schema.oaf.*;
import org.apache.commons.io.IOUtils;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.io.compress.CompressionCodec;
import org.apache.hadoop.io.compress.GzipCodec;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
@ -18,6 +21,9 @@ import java.util.*;
import static eu.dnetlib.dhp.PropagationConstant.*;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkHiveSession;
public class SparkCountryPropagationJob2 {
private static final Logger log = LoggerFactory.getLogger(SparkCountryPropagationJob2.class);
@ -35,30 +41,54 @@ public class SparkCountryPropagationJob2 {
parser.parseArgument(args);
Boolean isSparkSessionManaged = isSparkSessionManaged(parser);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
String inputPath = parser.get("sourcePath");
log.info("inputPath: {}", inputPath);
final String outputPath = "/tmp/provision/propagation/countrytoresultfrominstitutionalrepositories";
final String datasourcecountrypath = outputPath + "/prepared_datasource_country";
final String resultClassName = parser.get("resultTableName");
final String outputPath = parser.get("outputPath");
log.info("outputPath: {}", outputPath);
final String resultType = resultClassName.substring(resultClassName.lastIndexOf(".")+1);
final String datasourcecountrypath = parser.get("preparedInfoPath");
log.info("preparedInfoPath: {}", datasourcecountrypath);
final String resultClassName = parser.get("resultTableName");
log.info("resultTableName: {}", resultClassName);
final String resultType = resultClassName.substring(resultClassName.lastIndexOf(".") + 1).toLowerCase();
log.info("resultType: {}", resultType);
final Boolean writeUpdates = Optional
.ofNullable(parser.get("writeUpdate"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("writeUpdate: {}", writeUpdates);
final Boolean saveGraph = Optional
.ofNullable(parser.get("saveGraph"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("saveGraph: {}", saveGraph);
Class<? extends Result> resultClazz = (Class<? extends Result>) Class.forName(resultClassName);
SparkConf conf = new SparkConf();
conf.set("hive.metastore.uris", parser.get("hive_metastore_uris"));
final SparkSession spark = SparkSession
.builder()
.appName(SparkCountryPropagationJob2.class.getSimpleName())
.master(parser.get("master"))
.config(conf)
.enableHiveSupport()
.getOrCreate();
final boolean writeUpdates = TRUE.equals(parser.get("writeUpdate"));
final boolean saveGraph = TRUE.equals(parser.get("saveGraph"));
runWithSparkHiveSession(conf, isSparkSessionManaged,
spark -> {
//createOutputDirs(outputPath, FileSystem.get(spark.sparkContext().hadoopConfiguration()));
removeOutputDir(spark, outputPath);
execPropagation(spark, datasourcecountrypath, inputPath, outputPath, resultClazz, resultType,
writeUpdates, saveGraph);
});
}
private static <R extends Result> void execPropagation(SparkSession spark, String datasourcecountrypath,
String inputPath, String outputPath, Class<R> resultClazz, String resultType,
boolean writeUpdates, boolean saveGraph){
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
//Load parque file with preprocessed association datasource - country
@ -74,10 +104,9 @@ public class SparkCountryPropagationJob2 {
}
if(saveGraph){
updateResultTable(spark, potentialUpdates, inputPath, resultClazz, outputPath + "/" + resultType);
updateResultTable(spark, potentialUpdates, inputPath, resultClazz, outputPath);
}
}
private static <R extends Result> void updateResultTable(SparkSession spark, Dataset<ResultCountrySet> potentialUpdates,
@ -96,33 +125,32 @@ public class SparkCountryPropagationJob2 {
Encoders.tuple(Encoders.STRING(), Encoders.bean(ResultCountrySet.class)));
Dataset<R> new_table = result_pair
.joinWith(potential_update_pair, result_pair.col("_1").equalTo(potential_update_pair.col("_1")), "left")
.map((MapFunction<Tuple2<Tuple2<String, R>, Tuple2<String, ResultCountrySet>>, R>) value -> {
.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()).map(Tuple2::_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());
}
for (Country country : potentialNewCountries.get().getCountrySet()) {
for (CountrySbs country : potentialNewCountries.get().getCountrySet()) {
if (!countries.contains(country.getClassid())) {
r.getCountry().add(getCountry(country.getClassid(), country.getClassname()));
}
}
}
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);
.toJavaRDD()
.map(r -> OBJECT_MAPPER.writeValueAsString(r))
.saveAsTextFile(outputPath , GzipCodec.class);
}
@ -133,6 +161,7 @@ public class SparkCountryPropagationJob2 {
Dataset<R> result = readPathEntity(spark, inputPath, resultClazz);
result.createOrReplaceTempView("result");
//log.info("number of results: {}", result.count());
createCfHbforresult(spark);
return countryPropagationAssoc(spark, broadcast_datasourcecountryassoc);
}
@ -147,6 +176,7 @@ public class SparkCountryPropagationJob2 {
"LATERAL VIEW EXPLODE(instance) i AS inst";
Dataset<Row> cfhb = spark.sql(query);
cfhb.createOrReplaceTempView("cfhb");
log.info("cfhb_number : {}", cfhb.count());
}
@ -155,19 +185,22 @@ public class SparkCountryPropagationJob2 {
Dataset<DatasourceCountry> datasource_country = broadcast_datasourcecountryassoc.value();
datasource_country.createOrReplaceTempView("datasource_country");
log.info("datasource_country number : {}",datasource_country.count());
String query = "SELECT id, collect_set(country) country "+
String query = "SELECT id resultId, collect_set(country) countrySet "+
"FROM ( SELECT id, country " +
"FROM rels " +
"FROM datasource_country " +
"JOIN cfhb " +
" ON cf = ds " +
" ON cf = dataSourceId " +
"UNION ALL " +
"SELECT id , country " +
"FROM rels " +
"FROM datasource_country " +
"JOIN cfhb " +
" ON hb = ds ) tmp " +
" ON hb = dataSourceId ) tmp " +
"GROUP BY id";
return spark.sql(query);
Dataset<Row> potentialUpdates = spark.sql(query);
log.info("potential update number : {}", potentialUpdates.count());
return potentialUpdates;
}
private static <R extends Result> Dataset<R> readPathEntity(SparkSession spark, String inputEntityPath, Class<R> resultClazz) {
@ -180,14 +213,15 @@ public class SparkCountryPropagationJob2 {
}
private static Dataset<DatasourceCountry> readAssocDatasourceCountry(SparkSession spark, String relationPath) {
return spark.read()
.load(relationPath)
.as(Encoders.bean(DatasourceCountry.class));
return spark
.read()
.textFile(relationPath)
.map(value -> OBJECT_MAPPER.readValue(value, DatasourceCountry.class), Encoders.bean(DatasourceCountry.class));
}
private static void writeUpdates(JavaRDD<ResultCountrySet> potentialUpdates, String outputPath){
potentialUpdates.map(u -> OBJECT_MAPPER.writeValueAsString(u))
.saveAsTextFile(outputPath);
.saveAsTextFile(outputPath, GzipCodec.class);
}