Merge branch '8172_impact_indicators_workflow' of https://code-repo.d4science.org/D-Net/dnet-hadoop into 8172_impact_indicators_workflow

This commit is contained in:
Ilias Kanellos 2023-04-28 13:23:49 +03:00
commit a98da54896
19 changed files with 474 additions and 98 deletions

View File

@ -9,6 +9,7 @@ import java.util.List;
import java.util.Optional; import java.util.Optional;
import java.util.stream.Collectors; import java.util.stream.Collectors;
import eu.dnetlib.dhp.actionmanager.bipmodel.score.deserializers.BipProjectModel;
import org.apache.commons.io.IOUtils; import org.apache.commons.io.IOUtils;
import org.apache.hadoop.io.Text; import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapred.SequenceFileOutputFormat; import org.apache.hadoop.mapred.SequenceFileOutputFormat;
@ -24,7 +25,7 @@ import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper; import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.bipmodel.BipDeserialize; import eu.dnetlib.dhp.actionmanager.bipmodel.score.deserializers.BipResultModel;
import eu.dnetlib.dhp.actionmanager.bipmodel.BipScore; import eu.dnetlib.dhp.actionmanager.bipmodel.BipScore;
import eu.dnetlib.dhp.application.ArgumentApplicationParser; import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.common.HdfsSupport; import eu.dnetlib.dhp.common.HdfsSupport;
@ -40,7 +41,8 @@ import scala.Tuple2;
*/ */
public class SparkAtomicActionScoreJob implements Serializable { public class SparkAtomicActionScoreJob implements Serializable {
private static final String DOI = "doi"; private static final String RESULT = "result";
private static final String PROJECT = "project";
private static final Logger log = LoggerFactory.getLogger(SparkAtomicActionScoreJob.class); private static final Logger log = LoggerFactory.getLogger(SparkAtomicActionScoreJob.class);
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper(); private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
@ -56,18 +58,17 @@ public class SparkAtomicActionScoreJob implements Serializable {
parser.parseArgument(args); parser.parseArgument(args);
Boolean isSparkSessionManaged = Optional Boolean isSparkSessionManaged = isSparkSessionManaged(parser);
.ofNullable(parser.get("isSparkSessionManaged"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged); log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
final String inputPath = parser.get("inputPath"); final String inputPath = parser.get("inputPath");
log.info("inputPath {}: ", inputPath); log.info("inputPath: {}", inputPath);
final String outputPath = parser.get("outputPath"); final String outputPath = parser.get("outputPath");
log.info("outputPath {}: ", outputPath); log.info("outputPath: {}", outputPath);
final String targetEntity = parser.get("targetEntity");
log.info("targetEntity: {}", targetEntity);
SparkConf conf = new SparkConf(); SparkConf conf = new SparkConf();
@ -76,17 +77,48 @@ public class SparkAtomicActionScoreJob implements Serializable {
isSparkSessionManaged, isSparkSessionManaged,
spark -> { spark -> {
removeOutputDir(spark, outputPath); removeOutputDir(spark, outputPath);
prepareResults(spark, inputPath, outputPath);
}); // follow different procedures for different target entities
switch (targetEntity) {
case RESULT:
prepareResults(spark, inputPath, outputPath);
break;
case PROJECT:
prepareProjects(spark, inputPath, outputPath);
break;
default:
throw new RuntimeException("Unknown target entity: " + targetEntity);
}
}
);
}
private static <I extends Project> void prepareProjects(SparkSession spark, String inputPath, String outputPath) {
// read input bip project scores
Dataset<BipProjectModel> projectScores = readPath(spark, inputPath, BipProjectModel.class);
projectScores.map( (MapFunction<BipProjectModel, Project>) bipProjectScores -> {
Project project = new Project();
project.setId(bipProjectScores.getProjectId());
project.setMeasures(bipProjectScores.toMeasures());
return project;
}, Encoders.bean(Project.class))
.toJavaRDD()
.map(p -> new AtomicAction(Project.class, p))
.mapToPair( aa -> new Tuple2<>(new Text(aa.getClazz().getCanonicalName()),
new Text(OBJECT_MAPPER.writeValueAsString(aa))))
.saveAsHadoopFile(outputPath, Text.class, Text.class, SequenceFileOutputFormat.class);
} }
private static <I extends Result> void prepareResults(SparkSession spark, String bipScorePath, String outputPath) { private static <I extends Result> void prepareResults(SparkSession spark, String bipScorePath, String outputPath) {
final JavaSparkContext sc = JavaSparkContext.fromSparkContext(spark.sparkContext()); final JavaSparkContext sc = JavaSparkContext.fromSparkContext(spark.sparkContext());
JavaRDD<BipDeserialize> bipDeserializeJavaRDD = sc JavaRDD<BipResultModel> bipDeserializeJavaRDD = sc
.textFile(bipScorePath) .textFile(bipScorePath)
.map(item -> OBJECT_MAPPER.readValue(item, BipDeserialize.class)); .map(item -> OBJECT_MAPPER.readValue(item, BipResultModel.class));
Dataset<BipScore> bipScores = spark Dataset<BipScore> bipScores = spark
.createDataset(bipDeserializeJavaRDD.flatMap(entry -> entry.keySet().stream().map(key -> { .createDataset(bipDeserializeJavaRDD.flatMap(entry -> entry.keySet().stream().map(key -> {
@ -159,12 +191,4 @@ public class SparkAtomicActionScoreJob implements Serializable {
HdfsSupport.remove(path, spark.sparkContext().hadoopConfiguration()); HdfsSupport.remove(path, spark.sparkContext().hadoopConfiguration());
} }
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));
}
} }

View File

@ -0,0 +1,69 @@
package eu.dnetlib.dhp.actionmanager.bipmodel.score.deserializers;
import com.opencsv.bean.CsvBindByPosition;
import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.oaf.KeyValue;
import eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils;
import lombok.AllArgsConstructor;
import lombok.Getter;
import lombok.NoArgsConstructor;
import lombok.Setter;
import eu.dnetlib.dhp.schema.oaf.Measure;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collections;
import java.util.List;
import static eu.dnetlib.dhp.actionmanager.Constants.*;
@NoArgsConstructor
@AllArgsConstructor
@Getter
@Setter
public class BipProjectModel {
String projectId;
String numOfInfluentialResults;
String numOfPopularResults;
String totalImpulse;
String totalCitationCount;
// each project bip measure has exactly one value, hence one key-value pair
private Measure createMeasure(String measureId, String measureValue) {
KeyValue kv = new KeyValue();
kv.setKey("score");
kv.setValue(measureValue);
kv.setDataInfo(
OafMapperUtils.dataInfo(
false,
UPDATE_DATA_INFO_TYPE,
true,
false,
OafMapperUtils.qualifier(
UPDATE_MEASURE_BIP_CLASS_ID,
UPDATE_CLASS_NAME,
ModelConstants.DNET_PROVENANCE_ACTIONS,
ModelConstants.DNET_PROVENANCE_ACTIONS),
"")
);
Measure measure = new Measure();
measure.setId(measureId);
measure.setUnit(Collections.singletonList(kv));
return measure;
}
public List<Measure> toMeasures() {
return Arrays.asList(
createMeasure("numOfInfluentialResults", numOfInfluentialResults),
createMeasure("numOfPopularResults", numOfPopularResults),
createMeasure("totalImpulse", totalImpulse),
createMeasure("totalCitationCount", totalCitationCount)
);
}
}

View File

@ -1,5 +1,7 @@
package eu.dnetlib.dhp.actionmanager.bipmodel; package eu.dnetlib.dhp.actionmanager.bipmodel.score.deserializers;
import eu.dnetlib.dhp.actionmanager.bipmodel.Score;
import java.io.Serializable; import java.io.Serializable;
import java.util.ArrayList; import java.util.ArrayList;
@ -11,9 +13,9 @@ import java.util.List;
* Only needed for deserialization purposes * Only needed for deserialization purposes
*/ */
public class BipDeserialize extends HashMap<String, List<Score>> implements Serializable { public class BipResultModel extends HashMap<String, List<Score>> implements Serializable {
public BipDeserialize() { public BipResultModel() {
super(); super();
} }

View File

@ -24,7 +24,7 @@ import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper; import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.bipmodel.BipDeserialize; import eu.dnetlib.dhp.actionmanager.bipmodel.score.deserializers.BipResultModel;
import eu.dnetlib.dhp.actionmanager.bipmodel.BipScore; import eu.dnetlib.dhp.actionmanager.bipmodel.BipScore;
import eu.dnetlib.dhp.application.ArgumentApplicationParser; import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.common.HdfsSupport; import eu.dnetlib.dhp.common.HdfsSupport;
@ -82,9 +82,9 @@ public class PrepareBipFinder implements Serializable {
final JavaSparkContext sc = JavaSparkContext.fromSparkContext(spark.sparkContext()); final JavaSparkContext sc = JavaSparkContext.fromSparkContext(spark.sparkContext());
JavaRDD<BipDeserialize> bipDeserializeJavaRDD = sc JavaRDD<BipResultModel> bipDeserializeJavaRDD = sc
.textFile(inputPath) .textFile(inputPath)
.map(item -> OBJECT_MAPPER.readValue(item, BipDeserialize.class)); .map(item -> OBJECT_MAPPER.readValue(item, BipResultModel.class));
spark spark
.createDataset(bipDeserializeJavaRDD.flatMap(entry -> entry.keySet().stream().map(key -> { .createDataset(bipDeserializeJavaRDD.flatMap(entry -> entry.keySet().stream().map(key -> {

View File

@ -16,5 +16,11 @@
"paramLongName": "outputPath", "paramLongName": "outputPath",
"paramDescription": "the path of the new ActionSet", "paramDescription": "the path of the new ActionSet",
"paramRequired": true "paramRequired": true
},
{
"paramName": "te",
"paramLongName": "targetEntity",
"paramDescription": "the type of target entity to be enriched; currently supported one of { 'result', 'project' }",
"paramRequired": true
} }
] ]

View File

@ -6,8 +6,9 @@ import static org.junit.jupiter.api.Assertions.*;
import java.io.IOException; import java.io.IOException;
import java.nio.file.Files; import java.nio.file.Files;
import java.nio.file.Path; import java.nio.file.Path;
import java.util.List;
import eu.dnetlib.dhp.schema.oaf.KeyValue;
import eu.dnetlib.dhp.schema.oaf.Project;
import org.apache.commons.io.FileUtils; import org.apache.commons.io.FileUtils;
import org.apache.hadoop.io.Text; import org.apache.hadoop.io.Text;
import org.apache.spark.SparkConf; import org.apache.spark.SparkConf;
@ -27,7 +28,6 @@ import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper; import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.schema.action.AtomicAction; import eu.dnetlib.dhp.schema.action.AtomicAction;
import eu.dnetlib.dhp.schema.oaf.Publication;
import eu.dnetlib.dhp.schema.oaf.Result; import eu.dnetlib.dhp.schema.oaf.Result;
public class SparkAtomicActionScoreJobTest { public class SparkAtomicActionScoreJobTest {
@ -37,8 +37,11 @@ public class SparkAtomicActionScoreJobTest {
private static SparkSession spark; private static SparkSession spark;
private static Path workingDir; private static Path workingDir;
private static final Logger log = LoggerFactory
.getLogger(SparkAtomicActionScoreJobTest.class); private final static String RESULT = "result";
private final static String PROJECT = "project";
private static final Logger log = LoggerFactory.getLogger(SparkAtomicActionScoreJobTest.class);
@BeforeAll @BeforeAll
public static void beforeAll() throws IOException { public static void beforeAll() throws IOException {
@ -69,29 +72,31 @@ public class SparkAtomicActionScoreJobTest {
spark.stop(); spark.stop();
} }
private void runJob(String inputPath, String outputPath, String targetEntity) throws Exception {
SparkAtomicActionScoreJob.main(
new String[] {
"-isSparkSessionManaged", Boolean.FALSE.toString(),
"-inputPath", inputPath,
"-outputPath", outputPath,
"-targetEntity", targetEntity,
}
);
}
@Test @Test
void testMatch() throws Exception { void testResultScores() throws Exception {
String bipScoresPath = getClass() final String targetEntity = RESULT;
.getResource("/eu/dnetlib/dhp/actionmanager/bipfinder/bip_scores_oid.json") String inputResultScores = getClass()
.getResource("/eu/dnetlib/dhp/actionmanager/bipfinder/result_bip_scores.json")
.getPath(); .getPath();
String outputPath = workingDir.toString() + "/" + targetEntity + "/actionSet";
SparkAtomicActionScoreJob // execute the job to generate the action sets for result scores
.main( runJob(inputResultScores, outputPath, targetEntity);
new String[] {
"-isSparkSessionManaged",
Boolean.FALSE.toString(),
"-inputPath",
bipScoresPath,
"-outputPath",
workingDir.toString() + "/actionSet"
});
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext()); final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
JavaRDD<Result> tmp = sc JavaRDD<Result> tmp = sc
.sequenceFile(workingDir.toString() + "/actionSet", Text.class, Text.class) .sequenceFile(outputPath, Text.class, Text.class)
.map(value -> OBJECT_MAPPER.readValue(value._2().toString(), AtomicAction.class)) .map(value -> OBJECT_MAPPER.readValue(value._2().toString(), AtomicAction.class))
.map(aa -> ((Result) aa.getPayload())); .map(aa -> ((Result) aa.getPayload()));
@ -140,4 +145,61 @@ public class SparkAtomicActionScoreJobTest {
} }
@Test
void testProjectScores() throws Exception {
String targetEntity = PROJECT;
String inputResultScores = getClass()
.getResource("/eu/dnetlib/dhp/actionmanager/bipfinder/project_bip_scores.json")
.getPath();
String outputPath = workingDir.toString() + "/" + targetEntity + "/actionSet";
// execute the job to generate the action sets for project scores
runJob(inputResultScores, outputPath, PROJECT);
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
JavaRDD<Project> projects = sc
.sequenceFile(outputPath, Text.class, Text.class)
.map(value -> OBJECT_MAPPER.readValue(value._2().toString(), AtomicAction.class))
.map(aa -> ((Project) aa.getPayload()));
// test the number of projects
assertEquals(4, projects.count());
String testProjectId = "40|nih_________::c02a8233e9b60f05bb418f0c9b714833";
// count that the project with id testProjectId is present
assertEquals(1, projects.filter(row -> row.getId().equals(testProjectId)).count());
projects.filter(row -> row.getId().equals(testProjectId))
.flatMap(r -> r.getMeasures().iterator())
.foreach(m -> {
log.info(m.getId() + " " + m.getUnit());
// ensure that only one score is present for each bip impact measure
assertEquals(1, m.getUnit().size());
KeyValue kv = m.getUnit().get(0);
// ensure that the correct key is provided, i.e. score
assertEquals("score", kv.getKey());
switch(m.getId()) {
case "numOfInfluentialResults":
assertEquals("0", kv.getValue());
break;
case "numOfPopularResults":
assertEquals("1", kv.getValue());
break;
case "totalImpulse":
assertEquals("25", kv.getValue());
break;
case "totalCitationCount":
assertEquals("43", kv.getValue());
break;
default:
fail("Unknown measure id in the context of projects");
}
});
}
} }

View File

@ -0,0 +1,4 @@
{"projectId":"40|nsf_________::d93e50d22374a1cf59f6a232413ea027","numOfInfluentialResults":0,"numOfPopularResults":10,"totalImpulse":181,"totalCitationCount":235}
{"projectId":"40|nih_________::1c93debc7085e440f245fbe70b2e8b21","numOfInfluentialResults":14,"numOfPopularResults":17,"totalImpulse":1558,"totalCitationCount":4226}
{"projectId":"40|nih_________::c02a8233e9b60f05bb418f0c9b714833","numOfInfluentialResults":0,"numOfPopularResults":1,"totalImpulse":25,"totalCitationCount":43}
{"projectId":"40|corda_______::d91dcf3a87dd7f72248fab0b8a4ba273","numOfInfluentialResults":2,"numOfPopularResults":3,"totalImpulse":78,"totalCitationCount":178}

View File

@ -1,4 +1,4 @@
# Ranking Workflow for Openaire Publications # Ranking Workflow for OpenAIRE Publications
This project contains the files for running a paper ranking workflow on the openaire graph using apache oozie. This project contains the files for running a paper ranking workflow on the openaire graph using apache oozie.
All scripts are written in python and the project setup follows the typical oozie workflow structure: All scripts are written in python and the project setup follows the typical oozie workflow structure:
@ -7,17 +7,15 @@ All scripts are written in python and the project setup follows the typical oozi
- a job.properties file specifying parameter values for the parameters used by the workflow - a job.properties file specifying parameter values for the parameters used by the workflow
- a set of python scripts used by the workflow - a set of python scripts used by the workflow
**NOTE**: the workflow depends on the external library of ranking scripts called BiP! Ranker. **NOTE**: the workflow depends on the external library of ranking scripts called [BiP! Ranker](https://github.com/athenarc/Bip-Ranker).
You can check out a specific tag/release of BIP! Ranker using maven, as described in the following section. You can check out a specific tag/release of BIP! Ranker using maven, as described in the following section.
## Check out a specific tag/release of BIP-Ranker ## Build and deploy
* Edit the `scmVersion` of the maven-scm-plugin in the pom.xml to point to the tag/release version you want to check out. Use the following command for packaging:
* Then, use maven to perform the checkout:
``` ```
mvn scm:checkout mvn package -Poozie-package -Dworkflow.source.dir=eu/dnetlib/dhp/oa/graph/impact_indicators -DskipTests
``` ```
* The code should be visible under `src/main/bip-ranker` folder. Note: edit the property `bip.ranker.tag` of the `pom.xml` file to specify the tag of [BIP-Ranker](https://github.com/athenarc/Bip-Ranker) that you want to use.

View File

@ -5,9 +5,8 @@
<modelVersion>4.0.0</modelVersion> <modelVersion>4.0.0</modelVersion>
<parent> <parent>
<groupId>eu.dnetlib.dhp</groupId> <groupId>eu.dnetlib.dhp</groupId>
<artifactId>dhp</artifactId> <artifactId>dhp-workflows</artifactId>
<version>1.2.5-SNAPSHOT</version> <version>1.2.5-SNAPSHOT</version>
<relativePath>../pom.xml</relativePath>
</parent> </parent>
<artifactId>dhp-impact-indicators</artifactId> <artifactId>dhp-impact-indicators</artifactId>
@ -16,6 +15,9 @@
<maven.compiler.source>8</maven.compiler.source> <maven.compiler.source>8</maven.compiler.source>
<maven.compiler.target>8</maven.compiler.target> <maven.compiler.target>8</maven.compiler.target>
<project.build.sourceEncoding>UTF-8</project.build.sourceEncoding> <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
<!-- Use this property to fetch a specific tag -->
<bip.ranker.tag>v1.0.0</bip.ranker.tag>
</properties> </properties>
<scm> <scm>
@ -32,10 +34,29 @@
<configuration> <configuration>
<connectionType>connection</connectionType> <connectionType>connection</connectionType>
<scmVersionType>tag</scmVersionType><!-- 'branch' can also be provided here --> <scmVersionType>tag</scmVersionType><!-- 'branch' can also be provided here -->
<scmVersion>v1.0.0</scmVersion><!-- in case of scmVersionType == 'branch', this field points to the branch name --> <scmVersion>${bip.ranker.tag}</scmVersion><!-- in case of scmVersionType == 'branch', this field points to the branch name -->
<checkoutDirectory>${project.build.directory}/../src/main/bip-ranker</checkoutDirectory> <checkoutDirectory>${project.build.directory}/${oozie.package.file.name}/${oozieAppDir}/bip-ranker</checkoutDirectory>
</configuration> </configuration>
<executions>
<execution>
<id>checkout-bip-ranker</id>
<phase>prepare-package</phase>
<goals>
<goal>checkout</goal>
</goals>
</execution>
</executions>
</plugin> </plugin>
</plugins> </plugins>
</build> </build>
<dependencies>
<dependency>
<groupId>eu.dnetlib.dhp</groupId>
<artifactId>dhp-aggregation</artifactId>
<version>${projectVersion}</version>
<scope>compile</scope>
</dependency>
</dependencies>
</project> </project>

View File

@ -90,3 +90,6 @@ oozie.wf.application.path=${wfAppPath}
# Path where the final output should be? # Path where the final output should be?
actionSetOutputPath=${workflowDataDir}/bip_actionsets/ actionSetOutputPath=${workflowDataDir}/bip_actionsets/
# The directory to store project impact indicators
projectImpactIndicatorsOutput=${workflowDataDir}/project_indicators

View File

@ -0,0 +1,108 @@
import sys
from pyspark.sql import SparkSession
from pyspark import SparkConf, SparkContext
import pyspark.sql.functions as F
from pyspark.sql.types import StringType, IntegerType, StructType, StructField
if len(sys.argv) < 8:
print("Usage: projects_impact.py <relations_folder> <influence_file> <popularity_file> <cc_file> <impulse_file> <num_partitions> <output_dir>")
sys.exit(-1)
appName = 'Project Impact Indicators'
conf = SparkConf().setAppName(appName)
sc = SparkContext(conf = conf)
spark = SparkSession.builder.appName(appName).getOrCreate()
sc.setLogLevel('OFF')
# input parameters
relations_fd = sys.argv[1]
influence_fd = sys.argv[2]
popularity_fd = sys.argv[3]
cc_fd = sys.argv[4]
impulse_fd = sys.argv[5]
num_partitions = int(sys.argv[6])
output_dir = sys.argv[7]
# schema for impact indicator files
impact_files_schema = StructType([
StructField('resultId', StringType(), False),
StructField('score', IntegerType(), False),
StructField('class', StringType(), False),
])
# list of impact indicators
impact_indicators = [
('influence', influence_fd, 'class'),
('popularity', popularity_fd, 'class'),
('impulse', impulse_fd, 'score'),
('citation_count', cc_fd, 'score')
]
'''
* Read impact indicator file and return a dataframe with the following schema:
* resultId: String
* indicator_name: Integer
'''
def read_df(fd, indicator_name, column_name):
return spark.read.schema(impact_files_schema)\
.option('delimiter', '\t')\
.option('header', False)\
.csv(fd)\
.select('resultId', F.col(column_name).alias(indicator_name))\
.repartition(num_partitions, 'resultId')
# Print dataframe schema, first 5 rows, and count
def print_df(df):
df.show(50)
df.printSchema()
print(df.count())
# Sets a null value to the column if the value is equal to the given value
def set_class_value_to_null(column, value):
return F.when(column != value, column).otherwise(F.lit(None))
# load and filter Project-to-Result relations
print("Reading relations")
relations = spark.read.json(relations_fd)\
.select(F.col('source').alias('projectId'), F.col('target').alias('resultId'), 'relClass', 'dataInfo.deletedbyinference', 'dataInfo.invisible')\
.where( (F.col('relClass') == 'produces') \
& (F.col('deletedbyinference') == "false")\
& (F.col('invisible') == "false"))\
.drop('deletedbyinference')\
.drop('invisible')\
.drop('relClass')\
.repartition(num_partitions, 'resultId')
for indicator_name, fd, column_name in impact_indicators:
print("Reading {} '{}' field from file".format(indicator_name, column_name))
df = read_df(fd, indicator_name, column_name)
# sets a zero value to the indicator column if the value is C5
if (column_name == 'class'):
df = df.withColumn(indicator_name, F.when(F.col(indicator_name).isin("C5"), 0).otherwise(1))
# print_df(df)
print("Joining {} to relations".format(indicator_name))
# NOTE: we use inner join because we want to keep only the results that have an impact score
# also note that all impact scores have the same set of results
relations = relations.join(df, 'resultId', 'inner')\
.repartition(num_partitions, 'resultId')
# uncomment to print non-null values count for each indicator
# for indicator_name, fd, column_name in impact_indicators:
# print("Counting non null values for {}".format(indicator_name))
# print(relations.filter(F.col(indicator_name).isNotNull()).count())
# sum the impact indicator values for each project
relations.groupBy('projectId')\
.agg(\
F.sum('influence').alias('numOfInfluentialResults'),\
F.sum('popularity').alias('numOfPopularResults'),\
F.sum('impulse').alias('totalImpulse'),\
F.sum('citation_count').alias('totalCitationCount')\
)\
.write.mode("overwrite")\
.json(output_dir, compression="gzip")

View File

@ -15,6 +15,8 @@
<case to="map-openaire-to-doi">${resume eq "map-ids"}</case> <case to="map-openaire-to-doi">${resume eq "map-ids"}</case>
<case to="map-scores-to-dois">${resume eq "map-scores"}</case> <case to="map-scores-to-dois">${resume eq "map-scores"}</case>
<case to="create-openaire-ranking-graph">${resume eq "start"}</case> <case to="create-openaire-ranking-graph">${resume eq "start"}</case>
<case to="project-impact-indicators">${resume eq "projects-impact"}</case>
<!-- TODO: add action set creation here --> <!-- TODO: add action set creation here -->
<default to="create-openaire-ranking-graph" /> <default to="create-openaire-ranking-graph" />
</switch> </switch>
@ -32,7 +34,6 @@
<prepare> <prepare>
<delete path="${synonymFolder}"/> <delete path="${synonymFolder}"/>
</prepare> </prepare>
<!-- using configs from an example on openaire --> <!-- using configs from an example on openaire -->
<master>yarn-cluster</master> <master>yarn-cluster</master>
@ -88,9 +89,8 @@
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties --> <!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
<name-node>${nameNode}</name-node> <name-node>${nameNode}</name-node>
<!-- using configs from an example on openaire -->
<!-- using configs from an example on openaire --> <master>yarn-cluster</master>
<master>yarn-cluster</master>
<mode>cluster</mode> <mode>cluster</mode>
<!-- This is the name of our job --> <!-- This is the name of our job -->
@ -129,7 +129,6 @@
<job-tracker>${jobTracker}</job-tracker> <job-tracker>${jobTracker}</job-tracker>
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties --> <!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
<name-node>${nameNode}</name-node> <name-node>${nameNode}</name-node>
<!-- using configs from an example on openaire --> <!-- using configs from an example on openaire -->
<master>yarn-cluster</master> <master>yarn-cluster</master>
@ -179,9 +178,8 @@
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties --> <!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
<name-node>${nameNode}</name-node> <name-node>${nameNode}</name-node>
<!-- using configs from an example on openaire -->
<!-- using configs from an example on openaire --> <master>yarn-cluster</master>
<master>yarn-cluster</master>
<mode>cluster</mode> <mode>cluster</mode>
<!-- This is the name of our job --> <!-- This is the name of our job -->
@ -233,7 +231,7 @@
<!-- Reference says: The master element indicates the url of the Spark Master. Ex: spark://host:port, mesos://host:port, yarn-cluster, yarn-master, or local. --> <!-- Reference says: The master element indicates the url of the Spark Master. Ex: spark://host:port, mesos://host:port, yarn-cluster, yarn-master, or local. -->
<!-- <master>local[*]</master> --> <!-- <master>local[*]</master> -->
<!-- Reference says: The mode element if present indicates the mode of spark, where to run spark driver program. Ex: client,cluster. | In my case I always have a client --> <!-- Reference says: The mode element if present indicates the mode of spark, where to run spark driver program. Ex: client,cluster. | In my case I always have a client -->
<!-- <mode>client</mode> --> <!-- <mode>client</mode> -->
<!-- using configs from an example on openaire --> <!-- using configs from an example on openaire -->
<master>yarn-cluster</master> <master>yarn-cluster</master>
@ -334,12 +332,12 @@
<!-- This should give the machine/root of the hdfs --> <!-- This should give the machine/root of the hdfs -->
<name-node>${nameNode}</name-node> <name-node>${nameNode}</name-node>
<!-- Exec is needed foor shell comands - points to type of shell command --> <!-- Exec is needed for shell commands - points to type of shell command -->
<exec>/usr/bin/bash</exec> <exec>/usr/bin/bash</exec>
<!-- name of script to run --> <!-- name of script to run -->
<argument>get_ranking_files.sh</argument> <argument>get_ranking_files.sh</argument>
<!-- We only pass the directory where we expect to find the rankings --> <!-- We only pass the directory where we expect to find the rankings -->
<argument>/${workflowDataDir}</argument> <argument>/${workflowDataDir}</argument>
<!-- the name of the file run --> <!-- the name of the file run -->
<file>${wfAppPath}/get_ranking_files.sh#get_ranking_files.sh</file> <file>${wfAppPath}/get_ranking_files.sh#get_ranking_files.sh</file>
@ -372,8 +370,8 @@
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties --> <!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
<name-node>${nameNode}</name-node> <name-node>${nameNode}</name-node>
<!-- using configs from an example on openaire --> <!-- using configs from an example on openaire -->
<master>yarn-cluster</master> <master>yarn-cluster</master>
<mode>cluster</mode> <mode>cluster</mode>
<!-- This is the name of our job --> <!-- This is the name of our job -->
@ -420,8 +418,8 @@
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties --> <!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
<name-node>${nameNode}</name-node> <name-node>${nameNode}</name-node>
<!-- using configs from an example on openaire --> <!-- using configs from an example on openaire -->
<master>yarn-cluster</master> <master>yarn-cluster</master>
<mode>cluster</mode> <mode>cluster</mode>
<!-- This is the name of our job --> <!-- This is the name of our job -->
@ -474,7 +472,6 @@
<prepare> <prepare>
<delete path="${synonymFolder}"/> <delete path="${synonymFolder}"/>
</prepare> </prepare>
<!-- using configs from an example on openaire --> <!-- using configs from an example on openaire -->
<master>yarn-cluster</master> <master>yarn-cluster</master>
@ -518,7 +515,6 @@
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties --> <!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
<name-node>${nameNode}</name-node> <name-node>${nameNode}</name-node>
<!-- using configs from an example on openaire --> <!-- using configs from an example on openaire -->
<master>yarn-cluster</master> <master>yarn-cluster</master>
<mode>cluster</mode> <mode>cluster</mode>
@ -558,21 +554,19 @@
</action> </action>
<action name="deleteOutputPathForActionSet"> <action name="deleteOutputPathForActionSet">
<fs> <fs>
<delete path="${actionSetOutputPath}"/> <delete path="${actionSetOutputPath}/results/"/>
<mkdir path="${actionSetOutputPath}"/> <delete path="${actionSetOutputPath}/projects/"/>
<!--
<delete path="${workingDir}"/> <mkdir path="${actionSetOutputPath}/results/"/>
<mkdir path="${workingDir}"/> <mkdir path="${actionSetOutputPath}/projects/"/>
--> </fs>
</fs> <ok to="createActionSetForResults"/>
<ok to="createActionSet"/>
<error to="actionset-delete-fail"/> <error to="actionset-delete-fail"/>
</action> </action>
<action name="createActionSetForResults">
<action name="createActionSet">
<spark xmlns="uri:oozie:spark-action:0.2"> <spark xmlns="uri:oozie:spark-action:0.2">
<master>yarn</master> <master>yarn</master>
<mode>cluster</mode> <mode>cluster</mode>
@ -590,13 +584,90 @@
--conf spark.sql.warehouse.dir=${sparkSqlWarehouseDir} --conf spark.sql.warehouse.dir=${sparkSqlWarehouseDir}
</spark-opts> </spark-opts>
<arg>--inputPath</arg><arg>${bipScorePath}</arg> <arg>--inputPath</arg><arg>${bipScorePath}</arg>
<arg>--outputPath</arg><arg>${actionSetOutputPath}</arg> <arg>--outputPath</arg><arg>${actionSetOutputPath}/results/</arg>
</spark> <arg>--targetEntity</arg><arg>result</arg>
<ok to="end"/> </spark>
<ok to="project-impact-indicators"/>
<error to="actionset-creation-fail"/> <error to="actionset-creation-fail"/>
</action> </action>
<action name="project-impact-indicators">
<!-- This is required as a tag for spark jobs, regardless of programming language -->
<spark xmlns="uri:oozie:spark-action:0.2">
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
<job-tracker>${jobTracker}</job-tracker>
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
<name-node>${nameNode}</name-node>
<!-- using configs from an example on openaire -->
<master>yarn-cluster</master>
<mode>cluster</mode>
<!-- This is the name of our job -->
<name>Project Impact Indicators</name>
<!-- Script name goes here -->
<jar>projects_impact.py</jar>
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
<spark-opts>--executor-memory 18G --executor-cores 4 --driver-memory 10G
--master yarn
--deploy-mode cluster
--conf spark.sql.shuffle.partitions=7680
--conf spark.extraListeners=${spark2ExtraListeners}
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
<!-- Script arguments here -->
<!-- graph data folder from which to read relations -->
<arg>${openaireDataInput}/relations</arg>
<!-- input files with impact indicators for results -->
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['pr_file']}</arg>
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['attrank_file']}</arg>
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['cc_file']}</arg>
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['impulse_file']}</arg>
<!-- number of partitions to be used on joins -->
<arg>7680</arg>
<arg>${projectImpactIndicatorsOutput}</arg>
<!-- This needs to point to the file on the hdfs i think -->
<file>${wfAppPath}/projects_impact.py#projects_impact.py</file>
</spark>
<!-- Do this after finishing okay -->
<ok to="createActionSetForProjects" />
<!-- Go there if we have an error -->
<error to="project-impact-indicators-fail" />
</action>
<action name="createActionSetForProjects">
<spark xmlns="uri:oozie:spark-action:0.2">
<master>yarn</master>
<mode>cluster</mode>
<name>Produces the atomic action with the bip finder scores for projects</name>
<class>eu.dnetlib.dhp.actionmanager.bipfinder.SparkAtomicActionScoreJob</class>
<jar>dhp-aggregation-${projectVersion}.jar</jar>
<spark-opts>
--executor-memory=${sparkExecutorMemory}
--executor-cores=${sparkExecutorCores}
--driver-memory=${sparkDriverMemory}
--conf spark.extraListeners=${spark2ExtraListeners}
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.sql.warehouse.dir=${sparkSqlWarehouseDir}
</spark-opts>
<arg>--inputPath</arg><arg>${projectImpactIndicatorsOutput}</arg>
<arg>--outputPath</arg><arg>${actionSetOutputPath}/projects/</arg>
<arg>--targetEntity</arg><arg>project</arg>
</spark>
<ok to="end"/>
<error to="actionset-project-creation-fail"/>
</action>
<!-- TODO: end the workflow--> <!-- TODO: end the workflow-->
@ -641,7 +712,14 @@
</kill> </kill>
<kill name="actionset-creation-fail"> <kill name="actionset-creation-fail">
<message>ActionSet creation failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message> <message>ActionSet creation for results failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
</kill> </kill>
<kill name="project-impact-indicators-fail">
<message>Calculating project impact indicators failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
</kill>
<kill name="actionset-project-creation-fail">
<message>ActionSet creation for projects failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
</kill>
</workflow-app> </workflow-app>

View File

@ -38,6 +38,7 @@
<module>dhp-usage-raw-data-update</module> <module>dhp-usage-raw-data-update</module>
<module>dhp-broker-events</module> <module>dhp-broker-events</module>
<module>dhp-doiboost</module> <module>dhp-doiboost</module>
<module>dhp-impact-indicators</module>
</modules> </modules>
<pluginRepositories> <pluginRepositories>