D-Net
/
dnet-hadoop
14
5
Fork 11
Code Issues 14 Pull Requests 13 Projects 2 Releases 2 Wiki Activity

master #11

Manually merged
claudio.atzori merged 275 commits from :master into enrichment_wfs 2020-05-11 15:14:56 +02:00
Conversation 1 Commits 275 Files Changed 156 +133 -233
2 changed files with 133 additions and 233 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files
Showing only changes of commit 03f7cb6402 - Show all commits

50
dhp-workflows/dhp-propagation/src/main/java/eu/dnetlib/dhp/countrypropagation/PrepareResultCountryAssociation.java
View File

@ -14,14 +14,13 @@ import org.apache.spark.sql.SaveMode;
import java.util.Arrays; import java.util.Arrays;
import java.util.List; import java.util.List;
import java.util.Optional;
import static eu.dnetlib.dhp.PropagationConstant.createOutputDirs; import static eu.dnetlib.dhp.PropagationConstant.createOutputDirs;
import static eu.dnetlib.dhp.PropagationConstant.getConstraintList; import static eu.dnetlib.dhp.PropagationConstant.getConstraintList;
import org.slf4j.Logger; import org.slf4j.Logger;
import org.slf4j.LoggerFactory; import org.slf4j.LoggerFactory;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession;
/** /**
* For the association of the country to the datasource * For the association of the country to the datasource
@ -38,53 +37,46 @@ public class PrepareResultCountryAssociation {
public static void main(String[] args) throws Exception { public static void main(String[] args) throws Exception {
String jsonConfiguration = IOUtils.toString(PrepareResultCountryAssociation.class String jsonConfiguration = IOUtils.toString(PrepareResultCountryAssociation.class
.getResourceAsStream("/eu/dnetlib/dhp/countrypropagation/input_countrypropagation_parameters.json")); .getResourceAsStream("/eu/dnetlib/dhp/countrypropagation/input_prepare_dc_assoc.json"));
final ArgumentApplicationParser parser = new ArgumentApplicationParser( final ArgumentApplicationParser parser = new ArgumentApplicationParser(
jsonConfiguration); jsonConfiguration);
parser.parseArgument(args); parser.parseArgument(args);
Boolean isSparkSessionManaged = Optional
.ofNullable(parser.get("isSparkSessionManaged"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
String inputPath = parser.get("sourcePath"); String inputPath = parser.get("sourcePath");
log.info("inputPath: {}", inputPath); log.info("inputPath: {}", inputPath);
String graphTableClassName = parser.get("graphTableClassName");
log.info("graphTableClassName: {}", graphTableClassName);
SparkConf conf = new SparkConf(); SparkConf conf = new SparkConf();
conf.set("hive.metastore.uris", parser.get("hive_metastore_uris")); conf.set("hive.metastore.uris", parser.get("hive_metastore_uris"));
runWithSparkSession(conf, isSparkSessionManaged,
spark -> {
removeOutputDir(spark, outputPath);
joinRelationEntity(spark, inputRelationsPath, inputEntityPath, entityClazz, outputPath);
});
final SparkSession spark = SparkSession final SparkSession spark = SparkSession
.builder() .builder()
.appName(PrepareResultCountryAssociation.class.getSimpleName()) .appName(SparkCountryPropagationJob.class.getSimpleName())
.master(parser.get("master")) .master(parser.get("master"))
.config(conf) .config(conf)
.enableHiveSupport() .enableHiveSupport()
.getOrCreate(); .getOrCreate();
//todo add link to working dir
final String outputPath = "/tmp/provision/propagation/countrytoresultfrominstitutionalrepositories"; final String outputPath = "/tmp/provision/propagation/countrytoresultfrominstitutionalrepositories";
createOutputDirs(outputPath, FileSystem.get(spark.sparkContext().hadoopConfiguration())); createOutputDirs(outputPath, FileSystem.get(spark.sparkContext().hadoopConfiguration()));
prepareDatasourceCountryAssociation(spark,
Arrays.asList(parser.get("whitelist").split(";")),
Arrays.asList(parser.get("allowedtypes").split(";")),
inputPath,
outputPath);
List<String> whitelist = Arrays.asList(parser.get("whitelist").split(";"));
List<String> allowedtypes = Arrays.asList(parser.get("allowedtypes").split(";"));
}
private static void prepareDatasourceCountryAssociation(SparkSession spark,
List<String> whitelist,
List<String> allowedtypes,
String inputPath,
String outputPath) {
String whitelisted = ""; String whitelisted = "";
for (String i : whitelist){ for (String i : whitelist){
whitelisted += " OR id = '" + i + "'"; whitelisted += " OR id = '" + i + "'";
@ -93,13 +85,13 @@ public class PrepareResultCountryAssociation {
Dataset<Datasource> datasource = spark.createDataset(sc.textFile(inputPath + "/datasource") Dataset<Datasource> datasource = spark.createDataset(sc.textFile(inputPath + "/datasource")
.map(item -> new ObjectMapper().readValue(item, Datasource.class)).rdd(), Encoders.bean(Datasource.class)); .map(item -> OBJECT_MAPPER.readValue(item, Datasource.class)).rdd(), Encoders.bean(Datasource.class));
Dataset<Relation> relation = spark.createDataset(sc.textFile(inputPath + "/relation") Dataset<Relation> relation = spark.createDataset(sc.textFile(inputPath + "/relation")
.map(item -> new ObjectMapper().readValue(item, Relation.class)).rdd(), Encoders.bean(Relation.class)); .map(item -> OBJECT_MAPPER.readValue(item, Relation.class)).rdd(), Encoders.bean(Relation.class));
Dataset<Organization> organization = spark.createDataset(sc.textFile(inputPath + "/organization") Dataset<Organization> organization = spark.createDataset(sc.textFile(inputPath + "/organization")
.map(item -> new ObjectMapper().readValue(item, Organization.class)).rdd(), Encoders.bean(Organization.class)); .map(item -> OBJECT_MAPPER.readValue(item, Organization.class)).rdd(), Encoders.bean(Organization.class));
datasource.createOrReplaceTempView("datasource"); datasource.createOrReplaceTempView("datasource");
relation.createOrReplaceTempView("relation"); relation.createOrReplaceTempView("relation");
@ -129,4 +121,6 @@ public class PrepareResultCountryAssociation {
} }
} }

290
dhp-workflows/dhp-propagation/src/main/java/eu/dnetlib/dhp/countrypropagation/SparkCountryPropagationJob2.java
View File

@ -4,21 +4,17 @@ import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.application.ArgumentApplicationParser; import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.schema.oaf.*; import eu.dnetlib.dhp.schema.oaf.*;
import org.apache.commons.io.IOUtils; import org.apache.commons.io.IOUtils;
import org.apache.hadoop.fs.FileSystem;
import org.apache.spark.SparkConf; import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaPairRDD; import org.apache.spark.api.java.JavaPairRDD;
import org.apache.spark.api.java.JavaRDD; import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext; import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.api.java.function.MapFunction;
import org.apache.spark.broadcast.Broadcast;
import org.apache.spark.sql.*;
import org.apache.spark.sql.Dataset; import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Encoders;
import org.apache.spark.sql.Row;
import org.apache.spark.sql.SparkSession;
import scala.Tuple2; import scala.Tuple2;
import java.util.ArrayList; import java.util.*;
import java.util.Arrays;
import java.util.HashSet;
import java.util.List;
import static eu.dnetlib.dhp.PropagationConstant.*; import static eu.dnetlib.dhp.PropagationConstant.*;
import org.slf4j.Logger; import org.slf4j.Logger;
@ -32,8 +28,25 @@ public class SparkCountryPropagationJob2 {
public static void main(String[] args) throws Exception { public static void main(String[] args) throws Exception {
final ArgumentApplicationParser parser = new ArgumentApplicationParser(IOUtils.toString(SparkCountryPropagationJob2.class.getResourceAsStream("/eu/dnetlib/dhp/countrypropagation/input_countrypropagation_parameters.json"))); String jsonConfiguration = IOUtils.toString(SparkCountryPropagationJob2.class
.getResourceAsStream("/eu/dnetlib/dhp/countrypropagation/input_countrypropagation_parameters.json"));
final ArgumentApplicationParser parser = new ArgumentApplicationParser(
jsonConfiguration);
parser.parseArgument(args); parser.parseArgument(args);
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("resultClazz");
final String resultType = resultClassName.substring(resultClassName.lastIndexOf(".")+1);
Class<? extends Result> resultClazz = (Class<? extends Result>) Class.forName(resultClassName);
SparkConf conf = new SparkConf(); SparkConf conf = new SparkConf();
conf.set("hive.metastore.uris", parser.get("hive_metastore_uris")); conf.set("hive.metastore.uris", parser.get("hive_metastore_uris"));
final SparkSession spark = SparkSession final SparkSession spark = SparkSession
@ -44,242 +57,135 @@ public class SparkCountryPropagationJob2 {
.enableHiveSupport() .enableHiveSupport()
.getOrCreate(); .getOrCreate();
final boolean writeUpdates = TRUE.equals(parser.get("writeUpdate"));
final String inputPath = parser.get("sourcePath"); final boolean saveGraph = TRUE.equals(parser.get("saveGraph"));
final String outputPath = "/tmp/provision/propagation/countrytoresultfrominstitutionalrepositories";
// createOutputDirs(outputPath, FileSystem.get(spark.sparkContext().hadoopConfiguration()));
boolean writeUpdates = TRUE.equals(parser.get("writeUpdate"));
boolean saveGraph = TRUE.equals(parser.get("saveGraph"));
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext()); final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
//Load parque 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);
Dataset<eu.dnetlib.dhp.schema.oaf.Dataset> dataset = spark.createDataset(sc.textFile(inputPath + "/dataset") Dataset<Row> potentialUpdates = getPotentialResultToUpdate(spark, inputPath, resultClazz, broadcast_datasourcecountryassoc);
.map(item -> new ObjectMapper().readValue(item, eu.dnetlib.dhp.schema.oaf.Dataset.class)).rdd(),
Encoders.bean(eu.dnetlib.dhp.schema.oaf.Dataset.class));
Dataset<OtherResearchProduct> other = spark.createDataset(sc.textFile(inputPath + "/otherresearchproduct")
.map(item -> new ObjectMapper().readValue(item, OtherResearchProduct.class)).rdd(),
Encoders.bean(OtherResearchProduct.class));
Dataset<Software> software = spark.createDataset(sc.textFile(inputPath + "/software")
.map(item -> new ObjectMapper().readValue(item, Software.class)).rdd(),
Encoders.bean(Software.class));
Dataset<Publication> publication = spark.createDataset(sc.textFile(inputPath + "/publication")
.map(item -> new ObjectMapper().readValue(item, Publication.class)).rdd(),
Encoders.bean(Publication.class));
//todo broadcast
software.createOrReplaceTempView("software");
final JavaRDD<Row> toupdateresultsoftware = propagateOnResult(spark, "software");
dataset.createOrReplaceTempView("dataset");
final JavaRDD<Row> toupdateresultdataset = propagateOnResult(spark, "dataset");
other.createOrReplaceTempView("other");
final JavaRDD<Row> toupdateresultother = propagateOnResult(spark, "other");
publication.createOrReplaceTempView("publication");
final JavaRDD<Row> toupdateresultpublication = propagateOnResult(spark, "publication");
if(writeUpdates){ if(writeUpdates){
writeUpdates(toupdateresultsoftware, toupdateresultdataset, toupdateresultother, toupdateresultpublication, outputPath); writeUpdates(potentialUpdates.toJavaRDD(), outputPath + "/update_" + resultType);
} }
if(saveGraph){ if(saveGraph){
createUpdateForSoftwareDataset(toupdateresultsoftware, inputPath, spark) updateResultTable(spark, potentialUpdates, inputPath, resultClazz, outputPath + "/" + resultType);
.map(s -> new ObjectMapper().writeValueAsString(s))
.saveAsTextFile(outputPath + "/software");
createUpdateForDatasetDataset(toupdateresultdataset,inputPath,spark)
.map(d -> new ObjectMapper().writeValueAsString(d))
.saveAsTextFile(outputPath + "/dataset");
createUpdateForOtherDataset(toupdateresultother, inputPath, spark)
.map(o -> new ObjectMapper().writeValueAsString(o))
.saveAsTextFile(outputPath + "/otherresearchproduct");
createUpdateForPublicationDataset(toupdateresultpublication, inputPath, spark)
.map(p -> new ObjectMapper().writeValueAsString(p))
.saveAsTextFile(outputPath + "/publication");
} }
} }
private static void writeUpdates(JavaRDD<Row> software, JavaRDD<Row> dataset, JavaRDD<Row> other , JavaRDD<Row> publication, String outputPath){ private static <R extends Result> void updateResultTable(SparkSession spark, Dataset<Row> potentialUpdates,
createUpdateForResultDatasetWrite(software, outputPath, "update_software"); String inputPath,
createUpdateForResultDatasetWrite(dataset, outputPath, "update_dataset"); Class<R> resultClazz,
createUpdateForResultDatasetWrite(other, outputPath, "update_other"); String outputPath) {
createUpdateForResultDatasetWrite(publication, outputPath, "update_publication");
}
private static JavaRDD<OtherResearchProduct> createUpdateForOtherDataset(JavaRDD<Row> toupdateresult, String inputPath, SparkSession spark) { Dataset<R> result = readPathEntity(spark, inputPath, resultClazz);
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
return sc.textFile(inputPath + "/otherresearchproduct") Dataset<Tuple2<String, R>> result_pair = result
.map(item -> new ObjectMapper().readValue(item, OtherResearchProduct.class)) .map(r -> new Tuple2<>(r.getId(), r),
.mapToPair(s -> new Tuple2<>(s.getId(), s)).leftOuterJoin(getStringResultJavaPairRDD(toupdateresult)) Encoders.tuple(Encoders.STRING(), Encoders.bean(resultClazz)));
.map(c -> {
OtherResearchProduct oaf = c._2()._1(); Dataset<Tuple2<String, List>> potential_update_pair = potentialUpdates.map(pu -> new Tuple2<>(pu.getString(0), pu.getList(1)),
List<Country> countryList = oaf.getCountry(); Encoders.tuple(Encoders.STRING(), Encoders.bean(List.class)));
if (c._2()._2().isPresent()) {
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, List>>, R>) value -> {
R r = value._1()._2();
Optional<List<Object>> potentialNewCountries = Optional.ofNullable(value._2()).map(Tuple2::_2);
if (potentialNewCountries != null) {
HashSet<String> countries = new HashSet<>(); HashSet<String> countries = new HashSet<>();
for (Qualifier country : countryList) { for (Qualifier country : r.getCountry()) {
countries.add(country.getClassid()); countries.add(country.getClassid());
} }
Result r = c._2()._2().get();
for (Country country : r.getCountry()) { for (Object country : potentialNewCountries.get()) {
if (!countries.contains(country.getClassid())) { if (!countries.contains(country)) {
countryList.add(country); r.getCountry().add(getCountry((String) country));
} }
} }
oaf.setCountry(countryList);
} }
return oaf; return r;
});
}, Encoders.bean(resultClazz));
log.info("Saving graph table to path: {}", outputPath);
result
.toJSON()
.write()
.option("compression", "gzip")
.text(outputPath);
} }
private static JavaRDD<Publication> createUpdateForPublicationDataset(JavaRDD<Row> toupdateresult, String inputPath, SparkSession spark) {
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
return sc.textFile(inputPath + "/publication")
.map(item -> new ObjectMapper().readValue(item, Publication.class)) private static <R extends Result> Dataset<Row> getPotentialResultToUpdate(SparkSession spark, String inputPath,
.mapToPair(s -> new Tuple2<>(s.getId(), s)).leftOuterJoin(getStringResultJavaPairRDD(toupdateresult)) Class<R> resultClazz,
.map(c -> { Broadcast<Dataset<DatasourceCountry>> broadcast_datasourcecountryassoc) {
Publication oaf = c._2()._1();
List<Country> countryList = oaf.getCountry(); Dataset<R> result = readPathEntity(spark, inputPath, resultClazz);
if (c._2()._2().isPresent()) { result.createOrReplaceTempView("result");
HashSet<String> countries = new HashSet<>(); createCfHbforresult(spark);
for (Qualifier country : countryList) { return countryPropagationAssoc(spark, broadcast_datasourcecountryassoc);
countries.add(country.getClassid());
}
Result r = c._2()._2().get();
for (Country country : r.getCountry()) {
if (!countries.contains(country.getClassid())) {
countryList.add(country);
}
}
oaf.setCountry(countryList);
}
return oaf;
});
} }
private static JavaRDD<Software> createUpdateForSoftwareDataset(JavaRDD<Row> toupdateresult, String inputPath, SparkSession spark) {
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
return sc.textFile(inputPath + "/software") private static void createCfHbforresult(SparkSession spark) {
.map(item -> new ObjectMapper().readValue(item, Software.class))
.mapToPair(s -> new Tuple2<>(s.getId(), s)).leftOuterJoin(getStringResultJavaPairRDD(toupdateresult))
.map(c -> {
Software oaf = c._2()._1();
List<Country> countryList = oaf.getCountry();
if (c._2()._2().isPresent()) {
HashSet<String> countries = new HashSet<>();
for (Qualifier country : countryList) {
countries.add(country.getClassid());
}
Result r = c._2()._2().get();
for (Country country : r.getCountry()) {
if (!countries.contains(country.getClassid())) {
countryList.add(country);
}
}
oaf.setCountry(countryList);
}
return oaf;
});
}
private static JavaRDD<eu.dnetlib.dhp.schema.oaf.Dataset> createUpdateForDatasetDataset(JavaRDD<Row> toupdateresult, String inputPath, SparkSession spark) {
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
return sc.textFile(inputPath + "/dataset")
.map(item -> new ObjectMapper().readValue(item, eu.dnetlib.dhp.schema.oaf.Dataset.class))
.mapToPair(d -> new Tuple2<>(d.getId(), d)).leftOuterJoin(getStringResultJavaPairRDD(toupdateresult))
.map(c -> {
eu.dnetlib.dhp.schema.oaf.Dataset oaf = c._2()._1();
List<Country> countryList = oaf.getCountry();
if (c._2()._2().isPresent()) {
HashSet<String> countries = new HashSet<>();
for (Qualifier country : countryList) {
countries.add(country.getClassid());
}
Result r = c._2()._2().get();
for (Country country : r.getCountry()) {
if (!countries.contains(country.getClassid())) {
countryList.add(country);
}
}
oaf.setCountry(countryList);
}
return oaf;
});
}
private static JavaRDD<Row> propagateOnResult(SparkSession spark, String result_type) {
String query; String query;
query = "SELECT id, inst.collectedfrom.key cf , inst.hostedby.key hb " + query = "SELECT id, inst.collectedfrom.key cf , inst.hostedby.key hb " +
"FROM ( SELECT id, instance " + "FROM ( SELECT id, instance " +
"FROM " + result_type + "FROM result " +
" WHERE datainfo.deletedbyinference = false) ds " + " WHERE datainfo.deletedbyinference = false) ds " +
"LATERAL VIEW EXPLODE(instance) i AS inst"; "LATERAL VIEW EXPLODE(instance) i AS inst";
Dataset<Row> cfhb = spark.sql(query); Dataset<Row> cfhb = spark.sql(query);
cfhb.createOrReplaceTempView("cfhb"); cfhb.createOrReplaceTempView("cfhb");
return countryPropagationAssoc(spark, "cfhb").toJavaRDD();
} }
private static Dataset<Row> countryPropagationAssoc(SparkSession spark, String cfhbTable){
private static Dataset<Row> countryPropagationAssoc(SparkSession spark,
Broadcast<Dataset<DatasourceCountry>> broadcast_datasourcecountryassoc){
Dataset<DatasourceCountry> datasource_country = broadcast_datasourcecountryassoc.value();
datasource_country.createOrReplaceTempView("datasource_country");
String query = "SELECT id, collect_set(country) country "+ String query = "SELECT id, collect_set(country) country "+
"FROM ( SELECT id, country " + "FROM ( SELECT id, country " +
"FROM rels " + "FROM rels " +
"JOIN " + cfhbTable + "JOIN cfhb " +
" ON cf = ds " + " ON cf = ds " +
"UNION ALL " + "UNION ALL " +
"SELECT id , country " + "SELECT id , country " +
"FROM rels " + "FROM rels " +
"JOIN " + cfhbTable + "JOIN cfhb " +
" ON hb = ds ) tmp " + " ON hb = ds ) tmp " +
"GROUP BY id"; "GROUP BY id";
return spark.sql(query); return spark.sql(query);
} }
private static JavaPairRDD<String, Result> getStringResultJavaPairRDD(JavaRDD<Row> toupdateresult) { private static <R extends Result> Dataset<R> readPathEntity(SparkSession spark, String inputEntityPath, Class<R> resultClazz) {
return toupdateresult.map(c -> {
List<Country> countryList = new ArrayList<>(); log.info("Reading Graph table from: {}", inputEntityPath);
List<String> tmp = c.getList(1); return spark
for (String country : tmp) { .read()
countryList.add(getCountry(country)); .textFile(inputEntityPath)
} .map((MapFunction<String, R>) value -> OBJECT_MAPPER.readValue(value, resultClazz), Encoders.bean(resultClazz));
Result r = new Result();
r.setId(c.getString(0));
r.setCountry(countryList);
return r;
}).mapToPair(r -> new Tuple2<>(r.getId(), r));
} }
private static void createUpdateForResultDatasetWrite(JavaRDD<Row> toupdateresult, String outputPath, String type){ private static Dataset<DatasourceCountry> readAssocDatasourceCountry(SparkSession spark, String relationPath) {
toupdateresult.map(c -> { return spark.read()
List<Country> countryList = new ArrayList<>(); .load(relationPath)
List<String> tmp = c.getList(1); .as(Encoders.bean(DatasourceCountry.class));
for (String country : tmp) {
countryList.add(getCountry(country));
} }
Result r = new Result();
r.setId(c.getString(0));
r.setCountry(countryList);
return r;
}).map(r ->new ObjectMapper().writeValueAsString(r)) private static void writeUpdates(JavaRDD<Row> potentialUpdates, String outputPath){
.saveAsTextFile(outputPath+"/"+type); potentialUpdates.map(u -> OBJECT_MAPPER.writeValueAsString(u))
.saveAsTextFile(outputPath);
} }