2020-05-18 13:04:06 +02:00
|
|
|
|
|
|
|
package eu.dnetlib.dhp.actionmanager.project;
|
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession;
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-28 10:07:00 +02:00
|
|
|
import java.util.*;
|
2020-05-18 13:04:06 +02:00
|
|
|
|
|
|
|
import org.apache.commons.io.IOUtils;
|
2020-05-19 18:42:50 +02:00
|
|
|
import org.apache.spark.SparkConf;
|
2020-05-28 17:29:53 +02:00
|
|
|
import org.apache.spark.api.java.function.FlatMapFunction;
|
2020-05-19 18:42:50 +02:00
|
|
|
import org.apache.spark.api.java.function.MapFunction;
|
|
|
|
import org.apache.spark.sql.Dataset;
|
|
|
|
import org.apache.spark.sql.Encoders;
|
2020-05-28 17:29:53 +02:00
|
|
|
import org.apache.spark.sql.SaveMode;
|
2020-05-19 18:42:50 +02:00
|
|
|
import org.apache.spark.sql.SparkSession;
|
|
|
|
import org.slf4j.Logger;
|
|
|
|
import org.slf4j.LoggerFactory;
|
2020-05-18 13:04:06 +02:00
|
|
|
|
|
|
|
import com.fasterxml.jackson.databind.ObjectMapper;
|
|
|
|
|
2020-09-28 12:13:33 +02:00
|
|
|
import eu.dnetlib.dhp.actionmanager.project.utils.CSVProject;
|
2020-05-18 13:04:06 +02:00
|
|
|
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
|
2020-05-19 18:42:50 +02:00
|
|
|
import eu.dnetlib.dhp.common.HdfsSupport;
|
|
|
|
import scala.Tuple2;
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-10-05 13:54:39 +02:00
|
|
|
/**
|
|
|
|
* Selects only the relevant information collected with the projects: project grant agreement, project programme code and
|
|
|
|
* project topic code for the projects that are also collected from OpenAIRE.
|
|
|
|
*/
|
2020-05-19 18:42:50 +02:00
|
|
|
public class PrepareProjects {
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2021-08-11 12:13:22 +02:00
|
|
|
private static final Logger log = LoggerFactory.getLogger(PrepareProjects.class);
|
2020-05-19 18:42:50 +02:00
|
|
|
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
public static void main(String[] args) throws Exception {
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
String jsonConfiguration = IOUtils
|
|
|
|
.toString(
|
|
|
|
PrepareProjects.class
|
|
|
|
.getResourceAsStream(
|
|
|
|
"/eu/dnetlib/dhp/actionmanager/project/prepare_project_parameters.json"));
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
final ArgumentApplicationParser parser = new ArgumentApplicationParser(jsonConfiguration);
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
parser.parseArgument(args);
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
Boolean isSparkSessionManaged = Optional
|
|
|
|
.ofNullable(parser.get("isSparkSessionManaged"))
|
|
|
|
.map(Boolean::valueOf)
|
|
|
|
.orElse(Boolean.TRUE);
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
final String projectPath = parser.get("projectPath");
|
|
|
|
log.info("projectPath {}: ", projectPath);
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
final String outputPath = parser.get("outputPath");
|
|
|
|
log.info("outputPath {}: ", outputPath);
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-28 23:50:32 +02:00
|
|
|
final String dbProjectPath = parser.get("dbProjectPath");
|
2020-05-28 17:29:53 +02:00
|
|
|
log.info("dbProjectPath {}: ", dbProjectPath);
|
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
SparkConf conf = new SparkConf();
|
2020-05-18 13:04:06 +02:00
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
runWithSparkSession(
|
|
|
|
conf,
|
|
|
|
isSparkSessionManaged,
|
|
|
|
spark -> {
|
|
|
|
removeOutputDir(spark, outputPath);
|
2020-05-28 17:29:53 +02:00
|
|
|
exec(spark, projectPath, dbProjectPath, outputPath);
|
2020-05-19 18:42:50 +02:00
|
|
|
});
|
2020-05-18 13:04:06 +02:00
|
|
|
}
|
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
private static void removeOutputDir(SparkSession spark, String path) {
|
|
|
|
HdfsSupport.remove(path, spark.sparkContext().hadoopConfiguration());
|
2020-05-18 13:04:06 +02:00
|
|
|
}
|
|
|
|
|
2020-05-28 23:50:32 +02:00
|
|
|
private static void exec(SparkSession spark, String projectPath, String dbProjectPath, String outputPath) {
|
|
|
|
Dataset<CSVProject> project = readPath(spark, projectPath, CSVProject.class);
|
2020-05-28 17:29:53 +02:00
|
|
|
Dataset<ProjectSubset> dbProjects = readPath(spark, dbProjectPath, ProjectSubset.class);
|
|
|
|
|
2020-05-28 23:50:32 +02:00
|
|
|
dbProjects
|
|
|
|
.joinWith(project, dbProjects.col("code").equalTo(project.col("id")), "left")
|
|
|
|
.flatMap(getTuple2CSVProjectFlatMapFunction(), Encoders.bean(CSVProject.class))
|
|
|
|
.filter(Objects::nonNull)
|
|
|
|
.write()
|
|
|
|
.mode(SaveMode.Overwrite)
|
|
|
|
.option("compression", "gzip")
|
|
|
|
.json(outputPath);
|
2020-05-18 13:04:06 +02:00
|
|
|
|
|
|
|
}
|
|
|
|
|
2020-05-28 23:50:32 +02:00
|
|
|
private static FlatMapFunction<Tuple2<ProjectSubset, CSVProject>, CSVProject> getTuple2CSVProjectFlatMapFunction() {
|
2021-05-14 10:58:12 +02:00
|
|
|
return value -> {
|
2020-05-28 23:50:32 +02:00
|
|
|
Optional<CSVProject> csvProject = Optional.ofNullable(value._2());
|
|
|
|
List<CSVProject> csvProjectList = new ArrayList<>();
|
|
|
|
if (csvProject.isPresent()) {
|
|
|
|
|
|
|
|
String[] programme = csvProject.get().getProgramme().split(";");
|
2020-09-28 17:20:11 +02:00
|
|
|
String topic = csvProject.get().getTopics();
|
2020-09-24 15:20:56 +02:00
|
|
|
|
2020-05-28 23:50:32 +02:00
|
|
|
Arrays
|
|
|
|
.stream(programme)
|
|
|
|
.forEach(p -> {
|
|
|
|
CSVProject proj = new CSVProject();
|
2020-09-28 17:20:11 +02:00
|
|
|
proj.setTopics(topic);
|
2020-09-24 15:20:56 +02:00
|
|
|
|
2020-05-28 23:50:32 +02:00
|
|
|
proj.setProgramme(p);
|
|
|
|
proj.setId(csvProject.get().getId());
|
|
|
|
csvProjectList.add(proj);
|
|
|
|
});
|
|
|
|
}
|
|
|
|
return csvProjectList.iterator();
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2020-05-19 18:42:50 +02:00
|
|
|
public static <R> Dataset<R> readPath(
|
|
|
|
SparkSession spark, String inputPath, Class<R> clazz) {
|
|
|
|
return spark
|
|
|
|
.read()
|
|
|
|
.textFile(inputPath)
|
|
|
|
.map((MapFunction<String, R>) value -> OBJECT_MAPPER.readValue(value, clazz), Encoders.bean(clazz));
|
2020-05-18 13:04:06 +02:00
|
|
|
}
|
|
|
|
}
|