Merge branch 'beta' into index_records

This commit is contained in:
Claudio Atzori 2024-03-25 16:12:49 +01:00
commit 82fc609c4f
114 changed files with 104594 additions and 1123 deletions

View File

@ -145,105 +145,6 @@ public class AuthorMerger {
return null;
}
/**
* This method tries to figure out when two author are the same in the contest
* of ORCID enrichment
*
* @param left Author in the OAF entity
* @param right Author ORCID
* @return based on a heuristic on the names of the authors if they are the same.
*/
public static boolean checkORCIDSimilarity(final Author left, final Author right) {
final Person pl = parse(left);
final Person pr = parse(right);
// If one of them didn't have a surname we verify if they have the fullName not empty
// and verify if the normalized version is equal
if (!(pl.getSurname() != null && pl.getSurname().stream().anyMatch(StringUtils::isNotBlank) &&
pr.getSurname() != null && pr.getSurname().stream().anyMatch(StringUtils::isNotBlank))) {
if (pl.getFullname() != null && !pl.getFullname().isEmpty() && pr.getFullname() != null
&& !pr.getFullname().isEmpty()) {
return pl
.getFullname()
.stream()
.anyMatch(
fl -> pr.getFullname().stream().anyMatch(fr -> normalize(fl).equalsIgnoreCase(normalize(fr))));
} else {
return false;
}
}
// The Authors have one surname in common
if (pl.getSurname().stream().anyMatch(sl -> pr.getSurname().stream().anyMatch(sr -> sr.equalsIgnoreCase(sl)))) {
// If one of them has only a surname and is the same we can say that they are the same author
if ((pl.getName() == null || pl.getName().stream().allMatch(StringUtils::isBlank)) ||
(pr.getName() == null || pr.getName().stream().allMatch(StringUtils::isBlank)))
return true;
// The authors have the same initials of Name in common
if (pl
.getName()
.stream()
.anyMatch(
nl -> pr
.getName()
.stream()
.anyMatch(nr -> nr.equalsIgnoreCase(nl))))
return true;
}
// Sometimes we noticed that publication have author wrote in inverse order Surname, Name
// We verify if we have an exact match between name and surname
if (pl.getSurname().stream().anyMatch(sl -> pr.getName().stream().anyMatch(nr -> nr.equalsIgnoreCase(sl))) &&
pl.getName().stream().anyMatch(nl -> pr.getSurname().stream().anyMatch(sr -> sr.equalsIgnoreCase(nl))))
return true;
else
return false;
}
//
/**
* Method to enrich ORCID information in one list of authors based on another list
*
* @param baseAuthor the Author List in the OAF Entity
* @param orcidAuthor The list of ORCID Author intersected
* @return The Author List of the OAF Entity enriched with the orcid Author
*/
public static List<Author> enrichOrcid(List<Author> baseAuthor, List<Author> orcidAuthor) {
if (baseAuthor == null || baseAuthor.isEmpty())
return orcidAuthor;
if (orcidAuthor == null || orcidAuthor.isEmpty())
return baseAuthor;
if (baseAuthor.size() == 1 && orcidAuthor.size() > 10)
return baseAuthor;
final List<Author> oAuthor = new ArrayList<>();
oAuthor.addAll(orcidAuthor);
baseAuthor.forEach(ba -> {
Optional<Author> aMatch = oAuthor.stream().filter(oa -> checkORCIDSimilarity(ba, oa)).findFirst();
if (aMatch.isPresent()) {
final Author sameAuthor = aMatch.get();
addPid(ba, sameAuthor.getPid());
oAuthor.remove(sameAuthor);
}
});
return baseAuthor;
}
private static void addPid(final Author a, final List<StructuredProperty> pids) {
if (a.getPid() == null) {
a.setPid(new ArrayList<>());
}
a.getPid().addAll(pids);
}
public static String pidToComparableString(StructuredProperty pid) {
final String classid = pid.getQualifier().getClassid() != null ? pid.getQualifier().getClassid().toLowerCase()
: "";

View File

@ -1,24 +1,6 @@
package eu.dnetlib.dhp.oa.merge;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession;
import static org.apache.spark.sql.functions.col;
import static org.apache.spark.sql.functions.when;
import java.util.Map;
import java.util.Optional;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.ForkJoinPool;
import java.util.stream.Collectors;
import org.apache.commons.io.IOUtils;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.function.MapFunction;
import org.apache.spark.api.java.function.ReduceFunction;
import org.apache.spark.sql.*;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.common.HdfsSupport;
import eu.dnetlib.dhp.common.vocabulary.VocabularyGroup;
@ -26,169 +8,186 @@ import eu.dnetlib.dhp.schema.common.EntityType;
import eu.dnetlib.dhp.schema.common.ModelSupport;
import eu.dnetlib.dhp.schema.oaf.OafEntity;
import eu.dnetlib.dhp.schema.oaf.utils.GraphCleaningFunctions;
import eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils;
import eu.dnetlib.dhp.schema.oaf.utils.MergeUtils;
import eu.dnetlib.dhp.utils.ISLookupClientFactory;
import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpException;
import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpService;
import org.apache.commons.io.IOUtils;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.function.MapFunction;
import org.apache.spark.api.java.function.ReduceFunction;
import org.apache.spark.sql.*;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import scala.Tuple2;
import java.util.Map;
import java.util.Optional;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.ForkJoinPool;
import java.util.stream.Collectors;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession;
import static org.apache.spark.sql.functions.col;
import static org.apache.spark.sql.functions.when;
/**
* Groups the graph content by entity identifier to ensure ID uniqueness
*/
public class GroupEntitiesSparkJob {
private static final Logger log = LoggerFactory.getLogger(GroupEntitiesSparkJob.class);
private static final Logger log = LoggerFactory.getLogger(GroupEntitiesSparkJob.class);
private static final Encoder<OafEntity> OAFENTITY_KRYO_ENC = Encoders.kryo(OafEntity.class);
private static final Encoder<OafEntity> OAFENTITY_KRYO_ENC = Encoders.kryo(OafEntity.class);
private ArgumentApplicationParser parser;
private ArgumentApplicationParser parser;
public GroupEntitiesSparkJob(ArgumentApplicationParser parser) {
this.parser = parser;
}
public GroupEntitiesSparkJob(ArgumentApplicationParser parser) {
this.parser = parser;
}
public static void main(String[] args) throws Exception {
public static void main(String[] args) throws Exception {
String jsonConfiguration = IOUtils
.toString(
GroupEntitiesSparkJob.class
.getResourceAsStream(
"/eu/dnetlib/dhp/oa/merge/group_graph_entities_parameters.json"));
final ArgumentApplicationParser parser = new ArgumentApplicationParser(jsonConfiguration);
parser.parseArgument(args);
String jsonConfiguration = IOUtils
.toString(
GroupEntitiesSparkJob.class
.getResourceAsStream(
"/eu/dnetlib/dhp/oa/merge/group_graph_entities_parameters.json"));
final ArgumentApplicationParser parser = new ArgumentApplicationParser(jsonConfiguration);
parser.parseArgument(args);
Boolean isSparkSessionManaged = Optional
.ofNullable(parser.get("isSparkSessionManaged"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
Boolean isSparkSessionManaged = Optional
.ofNullable(parser.get("isSparkSessionManaged"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
final String isLookupUrl = parser.get("isLookupUrl");
log.info("isLookupUrl: {}", isLookupUrl);
final String isLookupUrl = parser.get("isLookupUrl");
log.info("isLookupUrl: {}", isLookupUrl);
final ISLookUpService isLookupService = ISLookupClientFactory.getLookUpService(isLookupUrl);
final ISLookUpService isLookupService = ISLookupClientFactory.getLookUpService(isLookupUrl);
new GroupEntitiesSparkJob(parser).run(isSparkSessionManaged, isLookupService);
}
new GroupEntitiesSparkJob(parser).run(isSparkSessionManaged, isLookupService);
}
public void run(Boolean isSparkSessionManaged, ISLookUpService isLookUpService)
throws ISLookUpException {
public void run(Boolean isSparkSessionManaged, ISLookUpService isLookUpService)
throws ISLookUpException {
String graphInputPath = parser.get("graphInputPath");
log.info("graphInputPath: {}", graphInputPath);
String graphInputPath = parser.get("graphInputPath");
log.info("graphInputPath: {}", graphInputPath);
String checkpointPath = parser.get("checkpointPath");
log.info("checkpointPath: {}", checkpointPath);
String checkpointPath = parser.get("checkpointPath");
log.info("checkpointPath: {}", checkpointPath);
String outputPath = parser.get("outputPath");
log.info("outputPath: {}", outputPath);
String outputPath = parser.get("outputPath");
log.info("outputPath: {}", outputPath);
boolean filterInvisible = Boolean.parseBoolean(parser.get("filterInvisible"));
log.info("filterInvisible: {}", filterInvisible);
boolean filterInvisible = Boolean.parseBoolean(parser.get("filterInvisible"));
log.info("filterInvisible: {}", filterInvisible);
SparkConf conf = new SparkConf();
conf.set("spark.serializer", "org.apache.spark.serializer.KryoSerializer");
conf.registerKryoClasses(ModelSupport.getOafModelClasses());
SparkConf conf = new SparkConf();
conf.set("spark.serializer", "org.apache.spark.serializer.KryoSerializer");
conf.registerKryoClasses(ModelSupport.getOafModelClasses());
final VocabularyGroup vocs = VocabularyGroup.loadVocsFromIS(isLookUpService);
final VocabularyGroup vocs = VocabularyGroup.loadVocsFromIS(isLookUpService);
runWithSparkSession(
conf,
isSparkSessionManaged,
spark -> {
HdfsSupport.remove(checkpointPath, spark.sparkContext().hadoopConfiguration());
groupEntities(spark, graphInputPath, checkpointPath, outputPath, filterInvisible, vocs);
});
}
runWithSparkSession(
conf,
isSparkSessionManaged,
spark -> {
HdfsSupport.remove(checkpointPath, spark.sparkContext().hadoopConfiguration());
groupEntities(spark, graphInputPath, checkpointPath, outputPath, filterInvisible, vocs);
});
}
private static void groupEntities(
SparkSession spark,
String inputPath,
String checkpointPath,
String outputPath,
boolean filterInvisible, VocabularyGroup vocs) {
private static void groupEntities(
SparkSession spark,
String inputPath,
String checkpointPath,
String outputPath,
boolean filterInvisible, VocabularyGroup vocs) {
Dataset<OafEntity> allEntities = spark.emptyDataset(OAFENTITY_KRYO_ENC);
Dataset<OafEntity> allEntities = spark.emptyDataset(OAFENTITY_KRYO_ENC);
for (Map.Entry<EntityType, Class> e : ModelSupport.entityTypes.entrySet()) {
String entity = e.getKey().name();
Class<? extends OafEntity> entityClass = e.getValue();
String entityInputPath = inputPath + "/" + entity;
for (Map.Entry<EntityType, Class> e : ModelSupport.entityTypes.entrySet()) {
String entity = e.getKey().name();
Class<? extends OafEntity> entityClass = e.getValue();
String entityInputPath = inputPath + "/" + entity;
if (!HdfsSupport.exists(entityInputPath, spark.sparkContext().hadoopConfiguration())) {
continue;
}
if (!HdfsSupport.exists(entityInputPath, spark.sparkContext().hadoopConfiguration())) {
continue;
}
allEntities = allEntities
.union(
((Dataset<OafEntity>) spark
.read()
.schema(Encoders.bean(entityClass).schema())
.json(entityInputPath)
.filter("length(id) > 0")
.as(Encoders.bean(entityClass)))
.map((MapFunction<OafEntity, OafEntity>) r -> r, OAFENTITY_KRYO_ENC));
}
allEntities = allEntities
.union(
((Dataset<OafEntity>) spark
.read()
.schema(Encoders.bean(entityClass).schema())
.json(entityInputPath)
.filter("length(id) > 0")
.as(Encoders.bean(entityClass)))
.map((MapFunction<OafEntity, OafEntity>) r -> r, OAFENTITY_KRYO_ENC));
}
Dataset<?> groupedEntities = allEntities
.map(
(MapFunction<OafEntity, OafEntity>) entity -> GraphCleaningFunctions
.applyCoarVocabularies(entity, vocs),
OAFENTITY_KRYO_ENC)
.groupByKey((MapFunction<OafEntity, String>) OafEntity::getId, Encoders.STRING())
.reduceGroups((ReduceFunction<OafEntity>) OafMapperUtils::mergeEntities)
.map(
(MapFunction<Tuple2<String, OafEntity>, Tuple2<String, OafEntity>>) t -> new Tuple2<>(
t._2().getClass().getName(), t._2()),
Encoders.tuple(Encoders.STRING(), OAFENTITY_KRYO_ENC));
Dataset<?> groupedEntities = allEntities
.map(
(MapFunction<OafEntity, OafEntity>) entity -> GraphCleaningFunctions
.applyCoarVocabularies(entity, vocs),
OAFENTITY_KRYO_ENC)
.groupByKey((MapFunction<OafEntity, String>) OafEntity::getId, Encoders.STRING())
.reduceGroups((ReduceFunction<OafEntity>) MergeUtils::checkedMerge)
.map(
(MapFunction<Tuple2<String, OafEntity>, Tuple2<String, OafEntity>>) t -> new Tuple2<>(
t._2().getClass().getName(), t._2()),
Encoders.tuple(Encoders.STRING(), OAFENTITY_KRYO_ENC));
// pivot on "_1" (classname of the entity)
// created columns containing only entities of the same class
for (Map.Entry<EntityType, Class> e : ModelSupport.entityTypes.entrySet()) {
String entity = e.getKey().name();
Class<? extends OafEntity> entityClass = e.getValue();
// pivot on "_1" (classname of the entity)
// created columns containing only entities of the same class
for (Map.Entry<EntityType, Class> e : ModelSupport.entityTypes.entrySet()) {
String entity = e.getKey().name();
Class<? extends OafEntity> entityClass = e.getValue();
groupedEntities = groupedEntities
.withColumn(
entity,
when(col("_1").equalTo(entityClass.getName()), col("_2")));
}
groupedEntities = groupedEntities
.withColumn(
entity,
when(col("_1").equalTo(entityClass.getName()), col("_2")));
}
groupedEntities
.drop("_1", "_2")
.write()
.mode(SaveMode.Overwrite)
.option("compression", "gzip")
.save(checkpointPath);
groupedEntities
.drop("_1", "_2")
.write()
.mode(SaveMode.Overwrite)
.option("compression", "gzip")
.save(checkpointPath);
ForkJoinPool parPool = new ForkJoinPool(ModelSupport.entityTypes.size());
ForkJoinPool parPool = new ForkJoinPool(ModelSupport.entityTypes.size());
ModelSupport.entityTypes
.entrySet()
.stream()
.map(e -> parPool.submit(() -> {
String entity = e.getKey().name();
Class<? extends OafEntity> entityClass = e.getValue();
ModelSupport.entityTypes
.entrySet()
.stream()
.map(e -> parPool.submit(() -> {
String entity = e.getKey().name();
Class<? extends OafEntity> entityClass = e.getValue();
spark
.read()
.load(checkpointPath)
.select(col(entity).as("value"))
.filter("value IS NOT NULL")
.as(OAFENTITY_KRYO_ENC)
.map((MapFunction<OafEntity, OafEntity>) r -> r, (Encoder<OafEntity>) Encoders.bean(entityClass))
.filter(filterInvisible ? "dataInfo.invisible != TRUE" : "TRUE")
.write()
.mode(SaveMode.Overwrite)
.option("compression", "gzip")
.json(outputPath + "/" + entity);
}))
.collect(Collectors.toList())
.forEach(t -> {
try {
t.get();
} catch (InterruptedException | ExecutionException e) {
throw new RuntimeException(e);
}
});
}
spark
.read()
.load(checkpointPath)
.select(col(entity).as("value"))
.filter("value IS NOT NULL")
.as(OAFENTITY_KRYO_ENC)
.map((MapFunction<OafEntity, OafEntity>) r -> r, (Encoder<OafEntity>) Encoders.bean(entityClass))
.filter(filterInvisible ? "dataInfo.invisible != TRUE" : "TRUE")
.write()
.mode(SaveMode.Overwrite)
.option("compression", "gzip")
.json(outputPath + "/" + entity);
}))
.collect(Collectors.toList())
.forEach(t -> {
try {
t.get();
} catch (InterruptedException | ExecutionException e) {
throw new RuntimeException(e);
}
});
}
}

View File

@ -0,0 +1,79 @@
package eu.dnetlib.dhp.schema.oaf.utils;
//
// Source code recreated from a .class file by IntelliJ IDEA
// (powered by FernFlower decompiler)
//
import eu.dnetlib.dhp.schema.common.EntityType;
import eu.dnetlib.dhp.schema.oaf.KeyValue;
import eu.dnetlib.dhp.schema.oaf.Oaf;
import eu.dnetlib.dhp.schema.oaf.OafEntity;
import eu.dnetlib.dhp.schema.oaf.Result;
import java.util.Comparator;
import java.util.HashSet;
import java.util.Optional;
import java.util.stream.Collectors;
public class MergeComparator implements Comparator<Oaf> {
public MergeComparator() {
}
public int compare(Oaf left, Oaf right) {
// nulls at the end
if (left == null && right == null) {
return 0;
} else if (left == null) {
return -1;
} else if (right == null) {
return 1;
}
// invisible
if (left.getDataInfo() != null && left.getDataInfo().getInvisible() == true) {
if (right.getDataInfo() != null && right.getDataInfo().getInvisible() == false) {
return -1;
}
}
// collectedfrom
HashSet<String> lCf = getCollectedFromIds(left);
HashSet<String> rCf = getCollectedFromIds(right);
if (lCf.contains("10|openaire____::081b82f96300b6a6e3d282bad31cb6e2") && !rCf.contains("10|openaire____::081b82f96300b6a6e3d282bad31cb6e2")) {
return -1;
} else if (!lCf.contains("10|openaire____::081b82f96300b6a6e3d282bad31cb6e2") && rCf.contains("10|openaire____::081b82f96300b6a6e3d282bad31cb6e2")) {
return 1;
}
SubEntityType lClass = SubEntityType.fromClass(left.getClass());
SubEntityType rClass = SubEntityType.fromClass(right.getClass());
return lClass.ordinal() - rClass.ordinal();
}
protected HashSet<String> getCollectedFromIds(Oaf left) {
return (HashSet) Optional.ofNullable(left.getCollectedfrom()).map((cf) -> {
return (HashSet) cf.stream().map(KeyValue::getKey).collect(Collectors.toCollection(HashSet::new));
}).orElse(new HashSet());
}
enum SubEntityType {
publication, dataset, software, otherresearchproduct, datasource, organization, project;
/**
* Resolves the EntityType, given the relative class name
*
* @param clazz the given class name
* @param <T> actual OafEntity subclass
* @return the EntityType associated to the given class
*/
public static <T extends Oaf> SubEntityType fromClass(Class<T> clazz) {
return valueOf(clazz.getSimpleName().toLowerCase());
}
}
}

View File

@ -0,0 +1,707 @@
package eu.dnetlib.dhp.schema.oaf.utils;
import eu.dnetlib.dhp.schema.common.AccessRightComparator;
import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.common.ModelSupport;
import eu.dnetlib.dhp.schema.oaf.*;
import org.apache.commons.lang3.StringUtils;
import org.apache.commons.lang3.tuple.ImmutablePair;
import org.apache.commons.lang3.tuple.Pair;
import java.text.ParseException;
import java.util.*;
import java.util.stream.Collectors;
import java.util.stream.Stream;
import static com.google.common.base.Objects.firstNonNull;
import static com.google.common.base.Preconditions.checkArgument;
public class MergeUtils {
public static <T extends Oaf> T checkedMerge(final T left, final T right) {
return (T) merge(left, right, false);
}
public static Oaf merge(final Oaf left, final Oaf right) {
return merge(left, right, false);
}
public static Oaf merge(final Oaf left, final Oaf right, boolean checkDelegatedAuthority) {
if (sameClass(left, right, OafEntity.class)) {
return mergeEntities(left, right, checkDelegatedAuthority);
} else if (sameClass(left, right, Relation.class)) {
return mergeRelation((Relation) left, (Relation) right);
} else {
throw new RuntimeException(
String
.format(
"MERGE_FROM_AND_GET incompatible types: %s, %s",
left.getClass().getCanonicalName(), right.getClass().getCanonicalName()));
}
}
private static <T extends Oaf> boolean sameClass(Object left, Object right, Class<T> cls) {
return cls.isAssignableFrom(left.getClass()) && cls.isAssignableFrom(right.getClass());
}
private static Oaf mergeEntities(Oaf left, Oaf right, boolean checkDelegatedAuthority) {
if (sameClass(left, right, Result.class)) {
if (!left.getClass().equals(right.getClass()) || checkDelegatedAuthority) {
return mergeResultsOfDifferentTypes((Result)left, (Result) right);
}
if (sameClass(left, right, Publication.class)) {
return mergePublication((Publication) left, (Publication) right);
}
if (sameClass(left, right, Dataset.class)) {
return mergeDataset((Dataset) left, (Dataset) right);
}
if (sameClass(left, right, OtherResearchProduct.class)) {
return mergeORP((OtherResearchProduct) left, (OtherResearchProduct) right);
}
if (sameClass(left, right, Software.class)) {
return mergeSoftware((Software) left, (Software) right);
}
return mergeResult((Result) left, (Result) right);
} else if (sameClass(left, right, Datasource.class)) {
// TODO
final int trust = compareTrust(left, right);
return mergeOafEntityFields((Datasource) left, (Datasource) right, trust);
} else if (sameClass(left, right, Organization.class)) {
return mergeOrganization((Organization) left, (Organization) right);
} else if (sameClass(left, right, Project.class)) {
return mergeProject((Project) left, (Project) right);
} else {
throw new RuntimeException(
String
.format(
"MERGE_FROM_AND_GET incompatible types: %s, %s",
left.getClass().getCanonicalName(), right.getClass().getCanonicalName()));
}
}
/**
* This method is used in the global result grouping phase. It checks if one of the two is from a delegated authority
* https://graph.openaire.eu/docs/data-model/pids-and-identifiers#delegated-authorities and in that case it prefers
* such version.
* <p>
* Otherwise, it considers a resulttype priority order implemented in {@link ResultTypeComparator}
* and proceeds with the canonical property merging.
*
* @param left
* @param right
* @return
*/
private static <T extends Result> T mergeResultsOfDifferentTypes(T left, T right) {
final boolean leftFromDelegatedAuthority = isFromDelegatedAuthority(left);
final boolean rightFromDelegatedAuthority = isFromDelegatedAuthority(right);
if (leftFromDelegatedAuthority && !rightFromDelegatedAuthority) {
return left;
}
if (!leftFromDelegatedAuthority && rightFromDelegatedAuthority) {
return right;
}
//TODO: raise trust to have preferred fields from one or the other??
if (new ResultTypeComparator().compare(left, right) < 0) {
return mergeResult(left, right);
} else {
return mergeResult(right, left);
}
}
private static DataInfo chooseDataInfo(DataInfo left, DataInfo right, int trust) {
if (trust > 0) {
return left;
} else if (trust == 0) {
if (left == null || (left.getInvisible() != null && left.getInvisible().equals(Boolean.TRUE))) {
return right;
} else {
return left;
}
} else {
return right;
}
}
private static String chooseString(String left, String right, int trust) {
if (trust > 0) {
return left;
} else if (trust == 0) {
return StringUtils.isNotBlank(left) ? left : right;
} else {
return right;
}
}
private static <T> T chooseReference(T left, T right, int trust) {
if (trust > 0) {
return left;
} else if (trust == 0) {
return left != null ? left : right;
} else {
return right;
}
}
private static Long max(Long left, Long right) {
if (left == null)
return right;
if (right == null)
return left;
return Math.max(left, right);
}
// trust ??
private static Boolean booleanOR(Boolean a, Boolean b) {
if (a == null) {
return b;
} else if (b == null) {
return a;
}
return a || b;
}
private static <T> List<T> unionDistinctLists(final List<T> left, final List<T> right, int trust) {
if (left == null) {
return right;
} else if (right == null) {
return left;
}
List<T> h = trust >= 0 ? left : right;
List<T> l = trust >= 0 ? right : left;
return Stream.concat(h.stream(), l.stream())
.filter(Objects::nonNull)
.distinct()
.collect(Collectors.toList());
}
private static List<String> unionDistinctListOfString(final List<String> l, final List<String> r) {
if (l == null) {
return r;
} else if (r == null) {
return l;
}
return Stream.concat(l.stream(), r.stream())
.filter(StringUtils::isNotBlank)
.distinct()
.collect(Collectors.toList());
}
//TODO review
private static List<KeyValue> mergeKeyValue(List<KeyValue> left, List<KeyValue> right, int trust) {
if (trust < 0) {
List<KeyValue> s = left;
left = right;
right = s;
}
HashMap<String, KeyValue> values = new HashMap<>();
left.forEach(kv -> values.put(kv.getKey(), kv));
right.forEach(kv -> values.putIfAbsent(kv.getKey(), kv));
return new ArrayList<>(values.values());
}
private static List<StructuredProperty> unionTitle(List<StructuredProperty> left, List<StructuredProperty> right, int trust) {
if (left == null) {
return right;
} else if (right == null) {
return left;
}
List<StructuredProperty> h = trust >= 0 ? left : right;
List<StructuredProperty> l = trust >= 0 ? right : left;
return Stream.concat(h.stream(), l.stream())
.filter(Objects::isNull)
.distinct()
.collect(Collectors.toList());
}
/**
* Internal utility that merges the common OafEntity fields
*
* @param merged
* @param enrich
* @param <T>
* @return
*/
private static <T extends Oaf> T mergeOafFields(T merged, T enrich, int trust) {
//TODO: union of all values, but what does it mean with KeyValue pairs???
merged.setCollectedfrom(mergeKeyValue(merged.getCollectedfrom(), enrich.getCollectedfrom(), trust));
merged.setDataInfo(chooseDataInfo(merged.getDataInfo(), enrich.getDataInfo(), trust));
merged.setLastupdatetimestamp(max(merged.getLastupdatetimestamp(), enrich.getLastupdatetimestamp()));
return merged;
}
/**
* Internal utility that merges the common OafEntity fields
*
* @param original
* @param enrich
* @param <T>
* @return
*/
private static <T extends OafEntity> T mergeOafEntityFields(T original, T enrich, int trust) {
final T merged = mergeOafFields(original, enrich, trust);
merged.setOriginalId(unionDistinctListOfString(merged.getOriginalId(), enrich.getOriginalId()));
merged.setPid(unionDistinctLists(merged.getPid(), enrich.getPid(), trust));
// dateofcollection mettere today quando si fa merge
merged.setDateofcollection(chooseString(merged.getDateofcollection(), enrich.getDateofcollection(), trust));
// setDateoftransformation mettere vuota in dedup, nota per Claudio
merged.setDateoftransformation(chooseString(merged.getDateoftransformation(), enrich.getDateoftransformation(), trust));
// TODO: was missing in OafEntity.merge
merged.setExtraInfo(unionDistinctLists(merged.getExtraInfo(), enrich.getExtraInfo(), trust));
//oaiprovenanze da mettere a null quando si genera merge
merged.setOaiprovenance(chooseReference(merged.getOaiprovenance(), enrich.getOaiprovenance(), trust));
merged.setMeasures(unionDistinctLists(merged.getMeasures(), enrich.getMeasures(), trust));
return merged;
}
public static <T extends Relation> T mergeRelation(T original, T enrich) {
int trust = compareTrust(original, enrich);
T merge = mergeOafFields(original, enrich, trust);
checkArgument(Objects.equals(merge.getSource(), enrich.getSource()), "source ids must be equal");
checkArgument(Objects.equals(merge.getTarget(), enrich.getTarget()), "target ids must be equal");
checkArgument(Objects.equals(merge.getRelType(), enrich.getRelType()), "relType(s) must be equal");
checkArgument(
Objects.equals(merge.getSubRelType(), enrich.getSubRelType()), "subRelType(s) must be equal");
checkArgument(Objects.equals(merge.getRelClass(), enrich.getRelClass()), "relClass(es) must be equal");
//merge.setProvenance(mergeLists(merge.getProvenance(), enrich.getProvenance()));
//TODO: trust ??
merge.setValidated(booleanOR(merge.getValidated(), enrich.getValidated()));
try {
merge.setValidationDate(ModelSupport.oldest(merge.getValidationDate(), enrich.getValidationDate()));
} catch (ParseException e) {
throw new IllegalArgumentException(String
.format(
"invalid validation date format in relation [s:%s, t:%s]: %s", merge.getSource(),
merge.getTarget(),
merge.getValidationDate()));
}
// TODO keyvalue merge
merge.setProperties(mergeKeyValue(merge.getProperties(), enrich.getProperties(), trust));
return merge;
}
public static <T extends Result> T mergeResult(T original, T enrich) {
final int trust = compareTrust(original, enrich);
T merge = mergeOafEntityFields(original, enrich, trust);
if (merge.getProcessingchargeamount() == null || StringUtils.isBlank(merge.getProcessingchargeamount().getValue())) {
merge.setProcessingchargeamount(enrich.getProcessingchargeamount());
merge.setProcessingchargecurrency(enrich.getProcessingchargecurrency());
}
// author = usare la stessa logica che in dedup
merge.setAuthor(chooseReference(merge.getAuthor(), enrich.getAuthor(), trust));
// il primo che mi arriva secondo l'ordinamento per priorita'
merge.setResulttype(chooseReference(merge.getResulttype(), enrich.getResulttype(), trust));
// gestito come il resulttype perche' e' un subtype
merge.setMetaResourceType(chooseReference(merge.getMetaResourceType(), enrich.getMetaResourceType(), trust));
// spostiamo nell'instance e qui prendo il primo che arriva
merge.setLanguage(chooseReference(merge.getLanguage(), enrich.getLanguage(), trust));
// country lasicamo,o cosi' -> parentesi sul datainfo
merge.setCountry(unionDistinctLists(merge.getCountry(), enrich.getCountry(), trust));
//ok
merge.setSubject(unionDistinctLists(merge.getSubject(), enrich.getSubject(), trust));
// union per priority quindi vanno in append
merge.setTitle(unionTitle(merge.getTitle(), enrich.getTitle(), trust));
//ok
merge.setRelevantdate(unionDistinctLists(merge.getRelevantdate(), enrich.getRelevantdate(), trust));
// prima trust e poi longest list
merge.setDescription(longestLists(merge.getDescription(), enrich.getDescription()));
// trust piu' alto e poi piu' vecchia
merge.setDateofacceptance(chooseReference(merge.getDateofacceptance(), enrich.getDateofacceptance(), trust));
// ok, ma publisher va messo ripetibile
merge.setPublisher(chooseReference(merge.getPublisher(), enrich.getPublisher(), trust));
// ok
merge.setEmbargoenddate(chooseReference(merge.getEmbargoenddate(), enrich.getEmbargoenddate(), trust));
// ok
merge.setSource(unionDistinctLists(merge.getSource(), enrich.getSource(), trust));
// ok
merge.setFulltext(unionDistinctLists(merge.getFulltext(), enrich.getFulltext(), trust));
// ok
merge.setFormat(unionDistinctLists(merge.getFormat(), enrich.getFormat(), trust));
// ok
merge.setContributor(unionDistinctLists(merge.getContributor(), enrich.getContributor(), trust));
// prima prendo l'higher trust, su questo prendo il valore migliore nelle istanze TODO
// trust maggiore ma a parita' di trust il piu' specifico (base del vocabolario)
// vedi note
merge.setResourcetype(firstNonNull(merge.getResourcetype(), enrich.getResourcetype()));
// ok
merge.setCoverage(unionDistinctLists(merge.getCoverage(), enrich.getCoverage(), trust));
// most open ok
if (enrich.getBestaccessright() != null
&& new AccessRightComparator<>()
.compare(enrich.getBestaccessright(), merge.getBestaccessright()) < 0) {
merge.setBestaccessright(enrich.getBestaccessright());
}
// TODO merge of datainfo given same id
merge.setContext(unionDistinctLists(merge.getContext(), enrich.getContext(), trust));
//ok
merge.setExternalReference(unionDistinctLists(merge.getExternalReference(), enrich.getExternalReference(), trust));
//instance enrichment or union
// review instance equals => add pid to comparision
if (!isAnEnrichment(merge) && !isAnEnrichment(enrich))
merge.setInstance(unionDistinctLists(merge.getInstance(), enrich.getInstance(), trust));
else {
final List<Instance> enrichmentInstances = isAnEnrichment(merge) ? merge.getInstance()
: enrich.getInstance();
final List<Instance> enrichedInstances = isAnEnrichment(merge) ? enrich.getInstance()
: merge.getInstance();
if (isAnEnrichment(merge))
merge.setDataInfo(enrich.getDataInfo());
merge.setInstance(enrichInstances(enrichedInstances, enrichmentInstances));
}
merge.setEoscifguidelines(unionDistinctLists(merge.getEoscifguidelines(), enrich.getEoscifguidelines(), trust));
merge.setIsGreen(booleanOR(merge.getIsGreen(), enrich.getIsGreen()));
// OK but should be list of values
merge.setOpenAccessColor(chooseReference(merge.getOpenAccessColor(), enrich.getOpenAccessColor(), trust));
merge.setIsInDiamondJournal(booleanOR(merge.getIsInDiamondJournal(), enrich.getIsInDiamondJournal()));
merge.setPubliclyFunded(booleanOR(merge.getPubliclyFunded(), enrich.getPubliclyFunded()));
return merge;
}
private static <T extends OtherResearchProduct> T mergeORP(T original, T enrich) {
int trust = compareTrust(original, enrich);
final T merge = mergeResult(original, enrich);
merge.setContactperson(unionDistinctLists(merge.getContactperson(), enrich.getContactperson(), trust));
merge.setContactgroup(unionDistinctLists(merge.getContactgroup(), enrich.getContactgroup(), trust));
merge.setTool(unionDistinctLists(merge.getTool(), enrich.getTool(), trust));
return merge;
}
private static <T extends Software> T mergeSoftware(T original, T enrich) {
int trust = compareTrust(original, enrich);
final T merge = mergeResult(original, enrich);
merge.setDocumentationUrl(unionDistinctLists(merge.getDocumentationUrl(), enrich.getDocumentationUrl(), trust));
merge.setLicense(unionDistinctLists(merge.getLicense(), enrich.getLicense(), trust));
merge.setCodeRepositoryUrl(chooseReference(merge.getCodeRepositoryUrl(), enrich.getCodeRepositoryUrl(), trust));
merge.setProgrammingLanguage(chooseReference(merge.getProgrammingLanguage(), enrich.getProgrammingLanguage(), trust));
return merge;
}
private static <T extends Dataset> T mergeDataset(T original, T enrich) {
int trust = compareTrust(original, enrich);
T merge = mergeResult(original, enrich);
merge.setStoragedate(chooseReference(merge.getStoragedate(), enrich.getStoragedate(), trust));
merge.setDevice(chooseReference(merge.getDevice(), enrich.getDevice(), trust));
merge.setSize(chooseReference(merge.getSize(), enrich.getSize(), trust));
merge.setVersion(chooseReference(merge.getVersion(), enrich.getVersion(), trust));
merge.setLastmetadataupdate(chooseReference(merge.getLastmetadataupdate(), enrich.getLastmetadataupdate(), trust));
merge.setMetadataversionnumber(chooseReference(merge.getMetadataversionnumber(), enrich.getMetadataversionnumber(), trust));
merge.setGeolocation(unionDistinctLists(merge.getGeolocation(), enrich.getGeolocation(), trust));
return merge;
}
public static <T extends Publication> T mergePublication(T original, T enrich) {
final int trust = compareTrust(original, enrich);
T merged = mergeResult(original, enrich);
merged.setJournal(chooseReference(merged.getJournal(), enrich.getJournal(), trust));
return merged;
}
private static <T extends Organization> T mergeOrganization(T left, T enrich) {
int trust = compareTrust(left, enrich);
T merged = mergeOafEntityFields(left, enrich, trust);
merged.setLegalshortname(chooseReference(merged.getLegalshortname(), enrich.getLegalshortname(), trust));
merged.setLegalname(chooseReference(merged.getLegalname(), enrich.getLegalname(), trust));
merged.setAlternativeNames(unionDistinctLists(enrich.getAlternativeNames(), merged.getAlternativeNames(), trust));
merged.setWebsiteurl(chooseReference(merged.getWebsiteurl(), enrich.getWebsiteurl(), trust));
merged.setLogourl(chooseReference(merged.getLogourl(), enrich.getLogourl(), trust));
merged.setEclegalbody(chooseReference(merged.getEclegalbody(), enrich.getEclegalbody(), trust));
merged.setEclegalperson(chooseReference(merged.getEclegalperson(), enrich.getEclegalperson(), trust));
merged.setEcnonprofit(chooseReference(merged.getEcnonprofit(), enrich.getEcnonprofit(), trust));
merged.setEcresearchorganization(chooseReference(merged.getEcresearchorganization(), enrich.getEcresearchorganization(), trust));
merged.setEchighereducation(chooseReference(merged.getEchighereducation(), enrich.getEchighereducation(), trust));
merged.setEcinternationalorganizationeurinterests(chooseReference(merged.getEcinternationalorganizationeurinterests(), enrich.getEcinternationalorganizationeurinterests(), trust));
merged.setEcinternationalorganization(chooseReference(merged.getEcinternationalorganization(), enrich.getEcinternationalorganization(), trust));
merged.setEcenterprise(chooseReference(merged.getEcenterprise(), enrich.getEcenterprise(), trust));
merged.setEcsmevalidated(chooseReference(merged.getEcsmevalidated(), enrich.getEcsmevalidated(), trust));
merged.setEcnutscode(chooseReference(merged.getEcnutscode(), enrich.getEcnutscode(), trust));
merged.setCountry(chooseReference(merged.getCountry(), enrich.getCountry(), trust));
return merged;
}
public static <T extends Project> T mergeProject(T original, T enrich) {
int trust = compareTrust(original, enrich);
T merged = mergeOafEntityFields(original, enrich, trust);
merged.setWebsiteurl(chooseReference(merged.getWebsiteurl(), enrich.getWebsiteurl(), trust));
merged.setCode(chooseReference(merged.getCode(), enrich.getCode(), trust));
merged.setAcronym(chooseReference(merged.getAcronym(), enrich.getAcronym(), trust));
merged.setTitle(chooseReference(merged.getTitle(), enrich.getTitle(), trust));
merged.setStartdate(chooseReference(merged.getStartdate(), enrich.getStartdate(), trust));
merged.setEnddate(chooseReference(merged.getEnddate(), enrich.getEnddate(), trust));
merged.setCallidentifier(chooseReference(merged.getCallidentifier(), enrich.getCallidentifier(), trust));
merged.setKeywords(chooseReference(merged.getKeywords(), enrich.getKeywords(), trust));
merged.setDuration(chooseReference(merged.getDuration(), enrich.getDuration(), trust));
merged.setEcsc39(chooseReference(merged.getEcsc39(), enrich.getEcsc39(), trust));
merged.setOamandatepublications(chooseReference(merged.getOamandatepublications(), enrich.getOamandatepublications(), trust));
merged.setEcarticle29_3(chooseReference(merged.getEcarticle29_3(), enrich.getEcarticle29_3(), trust));
merged.setSubjects(unionDistinctLists(merged.getSubjects(), enrich.getSubjects(), trust));
merged.setFundingtree(unionDistinctLists(merged.getFundingtree(), enrich.getFundingtree(), trust));
merged.setContracttype(chooseReference(merged.getContracttype(), enrich.getContracttype(), trust));
merged.setOptional1(chooseReference(merged.getOptional1(), enrich.getOptional1(), trust));
merged.setOptional2(chooseReference(merged.getOptional2(), enrich.getOptional2(), trust));
merged.setJsonextrainfo(chooseReference(merged.getJsonextrainfo(), enrich.getJsonextrainfo(), trust));
merged.setContactfullname(chooseReference(merged.getContactfullname(), enrich.getContactfullname(), trust));
merged.setContactfax(chooseReference(merged.getContactfax(), enrich.getContactfax(), trust));
merged.setContactphone(chooseReference(merged.getContactphone(), enrich.getContactphone(), trust));
merged.setContactemail(chooseReference(merged.getContactemail(), enrich.getContactemail(), trust));
merged.setSummary(chooseReference(merged.getSummary(), enrich.getSummary(), trust));
merged.setCurrency(chooseReference(merged.getCurrency(), enrich.getCurrency(), trust));
//missin in Project.merge
merged.setTotalcost(chooseReference(merged.getTotalcost(), enrich.getTotalcost(), trust));
merged.setFundedamount(chooseReference(merged.getFundedamount(), enrich.getFundedamount(), trust));
// trust ??
if (enrich.getH2020topiccode() != null && StringUtils.isEmpty(merged.getH2020topiccode())) {
merged.setH2020topiccode(enrich.getH2020topiccode());
merged.setH2020topicdescription(enrich.getH2020topicdescription());
}
merged.setH2020classification(unionDistinctLists(merged.getH2020classification(), enrich.getH2020classification(), trust));
return merged;
}
/**
* Longest lists list.
*
* @param a the a
* @param b the b
* @return the list
*/
public static List<Field<String>> longestLists(List<Field<String>> a, List<Field<String>> b) {
if (a == null || b == null)
return a == null ? b : a;
return a.size() >= b.size() ? a : b;
}
/**
* This main method apply the enrichment of the instances
*
* @param toEnrichInstances the instances that could be enriched
* @param enrichmentInstances the enrichment instances
* @return list of instances possibly enriched
*/
private static List<Instance> enrichInstances(final List<Instance> toEnrichInstances,
final List<Instance> enrichmentInstances) {
final List<Instance> enrichmentResult = new ArrayList<>();
if (toEnrichInstances == null) {
return enrichmentResult;
}
if (enrichmentInstances == null) {
return enrichmentResult;
}
Map<String, Instance> ri = toInstanceMap(enrichmentInstances);
toEnrichInstances.forEach(i -> {
final List<Instance> e = findEnrichmentsByPID(i.getPid(), ri);
if (e != null && e.size() > 0) {
e.forEach(enr -> applyEnrichment(i, enr));
} else {
final List<Instance> a = findEnrichmentsByPID(i.getAlternateIdentifier(), ri);
if (a != null && a.size() > 0) {
a.forEach(enr -> applyEnrichment(i, enr));
}
}
enrichmentResult.add(i);
});
return enrichmentResult;
}
/**
* This method converts the list of instance enrichments
* into a Map where the key is the normalized identifier
* and the value is the instance itself
*
* @param ri the list of enrichment instances
* @return the result map
*/
private static Map<String, Instance> toInstanceMap(final List<Instance> ri) {
return ri
.stream()
.filter(i -> i.getPid() != null || i.getAlternateIdentifier() != null)
.flatMap(i -> {
final List<Pair<String, Instance>> result = new ArrayList<>();
if (i.getPid() != null)
i
.getPid()
.stream()
.filter(MergeUtils::validPid)
.forEach(p -> result.add(new ImmutablePair<>(extractKeyFromPid(p), i)));
if (i.getAlternateIdentifier() != null)
i
.getAlternateIdentifier()
.stream()
.filter(MergeUtils::validPid)
.forEach(p -> result.add(new ImmutablePair<>(extractKeyFromPid(p), i)));
return result.stream();
})
.collect(
Collectors
.toMap(
Pair::getLeft,
Pair::getRight,
(a, b) -> a));
}
private static boolean isFromDelegatedAuthority(Result r) {
return Optional
.ofNullable(r.getInstance())
.map(
instance -> instance
.stream()
.filter(i -> Objects.nonNull(i.getCollectedfrom()))
.map(i -> i.getCollectedfrom().getKey())
.anyMatch(cfId -> IdentifierFactory.delegatedAuthorityDatasourceIds().contains(cfId)))
.orElse(false);
}
/**
* Valid pid boolean.
*
* @param p the p
* @return the boolean
*/
private static boolean validPid(final StructuredProperty p) {
return p.getValue() != null && p.getQualifier() != null && p.getQualifier().getClassid() != null;
}
/**
* Normalize pid string.
*
* @param pid the pid
* @return the string
*/
private static String extractKeyFromPid(final StructuredProperty pid) {
if (pid == null)
return null;
final StructuredProperty normalizedPid = CleaningFunctions.normalizePidValue(pid);
return String.format("%s::%s", normalizedPid.getQualifier().getClassid(), normalizedPid.getValue());
}
/**
* This utility method finds the list of enrichment instances
* that match one or more PIDs in the input list
*
* @param pids the list of PIDs
* @param enrichments the List of enrichment instances having the same pid
* @return the list
*/
private static List<Instance> findEnrichmentsByPID(final List<StructuredProperty> pids,
final Map<String, Instance> enrichments) {
if (pids == null || enrichments == null)
return null;
return pids
.stream()
.map(MergeUtils::extractKeyFromPid)
.map(enrichments::get)
.filter(Objects::nonNull)
.collect(Collectors.toList());
}
/**
* Is an enrichment boolean.
*
* @param e the e
* @return the boolean
*/
private static boolean isAnEnrichment(OafEntity e) {
return e.getDataInfo() != null &&
e.getDataInfo().getProvenanceaction() != null
&& ModelConstants.PROVENANCE_ENRICH.equalsIgnoreCase(e.getDataInfo().getProvenanceaction().getClassid());
}
/**
* This method apply enrichment on a single instance
* The enrichment consists of replacing values on
* single attribute only if in the current instance is missing
* The only repeatable field enriched is measures
*
* @param merge the current instance
* @param enrichment the enrichment instance
*/
private static void applyEnrichment(final Instance merge, final Instance enrichment) {
if (merge == null || enrichment == null)
return;
merge.setLicense(firstNonNull(merge.getLicense(), enrichment.getLicense()));
merge.setAccessright(firstNonNull(merge.getAccessright(), enrichment.getAccessright()));
merge.setInstancetype(firstNonNull(merge.getInstancetype(), enrichment.getInstancetype()));
merge.setInstanceTypeMapping(firstNonNull(merge.getInstanceTypeMapping(), enrichment.getInstanceTypeMapping()));
merge.setHostedby(firstNonNull(merge.getHostedby(), enrichment.getHostedby()));
merge.setUrl(unionDistinctLists(merge.getUrl(), enrichment.getUrl(), 0));
merge.setDistributionlocation(firstNonNull(merge.getDistributionlocation(), enrichment.getDistributionlocation()));
merge.setCollectedfrom(firstNonNull(merge.getCollectedfrom(), enrichment.getCollectedfrom()));
// pid and alternateId are used for matching
merge.setDateofacceptance(firstNonNull(merge.getDateofacceptance(), enrichment.getDateofacceptance()));
merge.setProcessingchargeamount(firstNonNull(merge.getProcessingchargeamount(), enrichment.getProcessingchargeamount()));
merge.setProcessingchargecurrency(firstNonNull(merge.getProcessingchargecurrency(), enrichment.getProcessingchargecurrency()));
merge.setRefereed(firstNonNull(merge.getRefereed(), enrichment.getRefereed()));
merge.setMeasures(unionDistinctLists(merge.getMeasures(), enrichment.getMeasures(), 0));
merge.setFulltext(firstNonNull(merge.getFulltext(), enrichment.getFulltext()));
}
private static int compareTrust(Oaf a, Oaf b) {
String left = Optional
.ofNullable(a.getDataInfo())
.map(DataInfo::getTrust)
.orElse("0.0");
String right = Optional
.ofNullable(b.getDataInfo())
.map(DataInfo::getTrust)
.orElse("0.0");
return left.compareTo(right);
}
}

View File

@ -14,7 +14,6 @@ import java.util.stream.Collectors;
import org.apache.commons.lang3.StringUtils;
import eu.dnetlib.dhp.schema.common.AccessRightComparator;
import eu.dnetlib.dhp.schema.common.ModelSupport;
import eu.dnetlib.dhp.schema.oaf.*;
public class OafMapperUtils {
@ -22,65 +21,6 @@ public class OafMapperUtils {
private OafMapperUtils() {
}
public static Oaf merge(final Oaf left, final Oaf right) {
if (ModelSupport.isSubClass(left, OafEntity.class)) {
return mergeEntities((OafEntity) left, (OafEntity) right);
} else if (ModelSupport.isSubClass(left, Relation.class)) {
((Relation) left).mergeFrom((Relation) right);
} else {
throw new IllegalArgumentException("invalid Oaf type:" + left.getClass().getCanonicalName());
}
return left;
}
public static OafEntity mergeEntities(OafEntity left, OafEntity right) {
if (ModelSupport.isSubClass(left, Result.class)) {
return mergeResults((Result) left, (Result) right);
} else if (ModelSupport.isSubClass(left, Datasource.class)) {
left.mergeFrom(right);
} else if (ModelSupport.isSubClass(left, Organization.class)) {
left.mergeFrom(right);
} else if (ModelSupport.isSubClass(left, Project.class)) {
left.mergeFrom(right);
} else {
throw new IllegalArgumentException("invalid OafEntity subtype:" + left.getClass().getCanonicalName());
}
return left;
}
public static Result mergeResults(Result left, Result right) {
final boolean leftFromDelegatedAuthority = isFromDelegatedAuthority(left);
final boolean rightFromDelegatedAuthority = isFromDelegatedAuthority(right);
if (leftFromDelegatedAuthority && !rightFromDelegatedAuthority) {
return left;
}
if (!leftFromDelegatedAuthority && rightFromDelegatedAuthority) {
return right;
}
if (new ResultTypeComparator().compare(left, right) < 0) {
left.mergeFrom(right);
return left;
} else {
right.mergeFrom(left);
return right;
}
}
private static boolean isFromDelegatedAuthority(Result r) {
return Optional
.ofNullable(r.getInstance())
.map(
instance -> instance
.stream()
.filter(i -> Objects.nonNull(i.getCollectedfrom()))
.map(i -> i.getCollectedfrom().getKey())
.anyMatch(cfId -> IdentifierFactory.delegatedAuthorityDatasourceIds().contains(cfId)))
.orElse(false);
}
public static KeyValue keyValue(final String k, final String v) {
final KeyValue kv = new KeyValue();
kv.setKey(k);

View File

@ -0,0 +1,111 @@
package eu.dnetlib.dhp.schema.oaf.utils;
import static org.junit.jupiter.api.Assertions.*;
import static org.junit.jupiter.api.Assertions.assertEquals;
import java.io.IOException;
import java.util.HashSet;
import java.util.List;
import java.util.stream.Collectors;
import eu.dnetlib.dhp.schema.common.ModelSupport;
import org.apache.commons.io.IOUtils;
import org.junit.jupiter.api.Test;
import com.fasterxml.jackson.databind.DeserializationFeature;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.oaf.Dataset;
import eu.dnetlib.dhp.schema.oaf.KeyValue;
import eu.dnetlib.dhp.schema.oaf.Publication;
import eu.dnetlib.dhp.schema.oaf.Result;
public class MergeUtilsTest {
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper()
.configure(DeserializationFeature.FAIL_ON_UNKNOWN_PROPERTIES, false);
@Test
void testMergePubs() throws IOException {
Publication p1 = read("publication_1.json", Publication.class);
Publication p2 = read("publication_2.json", Publication.class);
Dataset d1 = read("dataset_1.json", Dataset.class);
Dataset d2 = read("dataset_2.json", Dataset.class);
assertEquals(1, p1.getCollectedfrom().size());
assertEquals(ModelConstants.CROSSREF_ID, p1.getCollectedfrom().get(0).getKey());
assertEquals(1, d2.getCollectedfrom().size());
assertFalse(cfId(d2.getCollectedfrom()).contains(ModelConstants.CROSSREF_ID));
assertEquals(1, p2.getCollectedfrom().size());
assertFalse(cfId(p2.getCollectedfrom()).contains(ModelConstants.CROSSREF_ID));
assertEquals(1, d1.getCollectedfrom().size());
assertTrue(cfId(d1.getCollectedfrom()).contains(ModelConstants.CROSSREF_ID));
final Result p1d2 = MergeUtils.checkedMerge(p1, d2);
assertEquals(ModelConstants.PUBLICATION_RESULTTYPE_CLASSID, p1d2.getResulttype().getClassid());
assertTrue(p1d2 instanceof Publication);
assertEquals(p1.getId(), p1d2.getId());
}
@Test
void testMergePubs_1() throws IOException {
Publication p2 = read("publication_2.json", Publication.class);
Dataset d1 = read("dataset_1.json", Dataset.class);
final Result p2d1 = MergeUtils.checkedMerge(p2, d1);
assertEquals((ModelConstants.DATASET_RESULTTYPE_CLASSID), p2d1.getResulttype().getClassid());
assertTrue(p2d1 instanceof Dataset);
assertEquals(d1.getId(), p2d1.getId());
assertEquals(2, p2d1.getCollectedfrom().size());
}
@Test
void testMergePubs_2() throws IOException {
Publication p1 = read("publication_1.json", Publication.class);
Publication p2 = read("publication_2.json", Publication.class);
Result p1p2 = MergeUtils.checkedMerge(p1, p2);
assertTrue(p1p2 instanceof Publication);
assertEquals(p1.getId(), p1p2.getId());
assertEquals(2, p1p2.getCollectedfrom().size());
}
@Test
void testDelegatedAuthority_1() throws IOException {
Dataset d1 = read("dataset_2.json", Dataset.class);
Dataset d2 = read("dataset_delegated.json", Dataset.class);
assertEquals(1, d2.getCollectedfrom().size());
assertTrue(cfId(d2.getCollectedfrom()).contains(ModelConstants.ZENODO_OD_ID));
Result res = (Result) MergeUtils.merge(d1, d2, true);
assertEquals(d2, res);
}
@Test
void testDelegatedAuthority_2() throws IOException {
Dataset p1 = read("publication_1.json", Dataset.class);
Dataset d2 = read("dataset_delegated.json", Dataset.class);
assertEquals(1, d2.getCollectedfrom().size());
assertTrue(cfId(d2.getCollectedfrom()).contains(ModelConstants.ZENODO_OD_ID));
Result res = (Result) MergeUtils.merge(p1, d2, true);
assertEquals(d2, res);
}
protected HashSet<String> cfId(List<KeyValue> collectedfrom) {
return collectedfrom.stream().map(KeyValue::getKey).collect(Collectors.toCollection(HashSet::new));
}
protected <T extends Result> T read(String filename, Class<T> clazz) throws IOException {
final String json = IOUtils.toString(getClass().getResourceAsStream(filename));
return OBJECT_MAPPER.readValue(json, clazz);
}
}

View File

@ -149,7 +149,7 @@ class OafMapperUtilsTest {
void testDate() {
final String date = GraphCleaningFunctions.cleanDate("23-FEB-1998");
assertNotNull(date);
System.out.println(date);
assertEquals("1998-02-23", date);
}
@Test
@ -166,8 +166,8 @@ class OafMapperUtilsTest {
assertEquals(
ModelConstants.PUBLICATION_RESULTTYPE_CLASSID,
OafMapperUtils
.mergeResults(p1, d2)
MergeUtils
.mergeResult(p1, d2)
.getResulttype()
.getClassid());
@ -178,8 +178,8 @@ class OafMapperUtilsTest {
assertEquals(
ModelConstants.DATASET_RESULTTYPE_CLASSID,
OafMapperUtils
.mergeResults(p2, d1)
MergeUtils
.mergeResult(p2, d1)
.getResulttype()
.getClassid());
}
@ -192,7 +192,7 @@ class OafMapperUtilsTest {
assertEquals(1, d2.getCollectedfrom().size());
assertTrue(cfId(d2.getCollectedfrom()).contains(ModelConstants.ZENODO_OD_ID));
Result res = OafMapperUtils.mergeResults(d1, d2);
Result res = MergeUtils.mergeResult(d1, d2);
assertEquals(d2, res);

View File

@ -1,14 +1,13 @@
package eu.dnetlib.dhp.actionmanager.promote;
import static eu.dnetlib.dhp.schema.common.ModelSupport.isSubClass;
import eu.dnetlib.dhp.common.FunctionalInterfaceSupport.SerializableSupplier;
import eu.dnetlib.dhp.schema.oaf.Oaf;
import eu.dnetlib.dhp.schema.oaf.utils.MergeUtils;
import java.util.function.BiFunction;
import eu.dnetlib.dhp.common.FunctionalInterfaceSupport.SerializableSupplier;
import eu.dnetlib.dhp.schema.oaf.Oaf;
import eu.dnetlib.dhp.schema.oaf.OafEntity;
import eu.dnetlib.dhp.schema.oaf.Relation;
import static eu.dnetlib.dhp.schema.common.ModelSupport.isSubClass;
/** OAF model merging support. */
public class MergeAndGet {
@ -46,20 +45,7 @@ public class MergeAndGet {
}
private static <G extends Oaf, A extends Oaf> G mergeFromAndGet(G x, A y) {
if (isSubClass(x, Relation.class) && isSubClass(y, Relation.class)) {
((Relation) x).mergeFrom((Relation) y);
return x;
} else if (isSubClass(x, OafEntity.class)
&& isSubClass(y, OafEntity.class)
&& isSubClass(x, y)) {
((OafEntity) x).mergeFrom((OafEntity) y);
return x;
}
throw new RuntimeException(
String
.format(
"MERGE_FROM_AND_GET incompatible types: %s, %s",
x.getClass().getCanonicalName(), y.getClass().getCanonicalName()));
return (G) MergeUtils.merge(x, y);
}
@SuppressWarnings("unchecked")

View File

@ -64,6 +64,9 @@ public class PrepareAffiliationRelations implements Serializable {
final String pubmedInputPath = parser.get("pubmedInputPath");
log.info("pubmedInputPath: {}", pubmedInputPath);
final String openapcInputPath = parser.get("openapcInputPath");
log.info("openapcInputPath: {}", openapcInputPath);
final String outputPath = parser.get("outputPath");
log.info("outputPath: {}", outputPath);
@ -85,8 +88,14 @@ public class PrepareAffiliationRelations implements Serializable {
JavaPairRDD<Text, Text> pubmedRelations = prepareAffiliationRelations(
spark, pubmedInputPath, collectedFromPubmed);
List<KeyValue> collectedFromOpenAPC = OafMapperUtils
.listKeyValues(ModelConstants.OPEN_APC_ID, "OpenAPC");
JavaPairRDD<Text, Text> openAPCRelations = prepareAffiliationRelations(
spark, openapcInputPath, collectedFromOpenAPC);
crossrefRelations
.union(pubmedRelations)
.union(openAPCRelations)
.saveAsHadoopFile(
outputPath, Text.class, Text.class, SequenceFileOutputFormat.class, GzipCodec.class);

View File

@ -95,7 +95,7 @@ public class SparkAtomicActionScoreJob implements Serializable {
return projectScores.map((MapFunction<BipProjectModel, Project>) bipProjectScores -> {
Project project = new Project();
project.setId(bipProjectScores.getProjectId());
//project.setId(bipProjectScores.getProjectId());
project.setMeasures(bipProjectScores.toMeasures());
return project;
}, Encoders.bean(Project.class))

View File

@ -34,6 +34,11 @@ public class BipProjectModel {
String totalCitationCount;
public String getProjectId() {
return projectId;
}
// each project bip measure has exactly one value, hence one key-value pair
private Measure createMeasure(String measureId, String measureValue) {

View File

@ -75,6 +75,7 @@ public class GetFOSSparkJob implements Serializable {
fosData.map((MapFunction<Row, FOSDataModel>) r -> {
FOSDataModel fosDataModel = new FOSDataModel();
fosDataModel.setDoi(r.getString(0).toLowerCase());
fosDataModel.setOaid(r.getString(1).toLowerCase());
fosDataModel.setLevel1(r.getString(2));
fosDataModel.setLevel2(r.getString(3));
fosDataModel.setLevel3(r.getString(4));

View File

@ -16,12 +16,14 @@ import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Encoders;
import org.apache.spark.sql.SaveMode;
import org.apache.spark.sql.SparkSession;
import org.jetbrains.annotations.NotNull;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import eu.dnetlib.dhp.actionmanager.createunresolvedentities.model.FOSDataModel;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.common.ModelSupport;
import eu.dnetlib.dhp.schema.oaf.Result;
import eu.dnetlib.dhp.schema.oaf.StructuredProperty;
import eu.dnetlib.dhp.schema.oaf.Subject;
@ -52,62 +54,90 @@ public class PrepareFOSSparkJob implements Serializable {
final String outputPath = parser.get("outputPath");
log.info("outputPath: {}", outputPath);
final Boolean distributeDOI = Optional
.ofNullable(parser.get("distributeDoi"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
SparkConf conf = new SparkConf();
runWithSparkSession(
conf,
isSparkSessionManaged,
spark -> {
distributeFOSdois(
spark,
sourcePath,
if (distributeDOI)
distributeFOSdois(
spark,
sourcePath,
outputPath);
outputPath);
else
distributeFOSoaid(spark, sourcePath, outputPath);
});
}
private static void distributeFOSoaid(SparkSession spark, String sourcePath, String outputPath) {
Dataset<FOSDataModel> fosDataset = readPath(spark, sourcePath, FOSDataModel.class);
fosDataset
.groupByKey((MapFunction<FOSDataModel, String>) v -> v.getOaid().toLowerCase(), Encoders.STRING())
.mapGroups((MapGroupsFunction<String, FOSDataModel, Result>) (k, it) -> {
return getResult(ModelSupport.getIdPrefix(Result.class) + "|" + k, it);
}, Encoders.bean(Result.class))
.write()
.mode(SaveMode.Overwrite)
.option("compression", "gzip")
.json(outputPath + "/fos");
}
@NotNull
private static Result getResult(String k, Iterator<FOSDataModel> it) {
Result r = new Result();
FOSDataModel first = it.next();
r.setId(k);
HashSet<String> level1 = new HashSet<>();
HashSet<String> level2 = new HashSet<>();
HashSet<String> level3 = new HashSet<>();
HashSet<String> level4 = new HashSet<>();
addLevels(level1, level2, level3, level4, first);
it.forEachRemaining(v -> addLevels(level1, level2, level3, level4, v));
List<Subject> sbjs = new ArrayList<>();
level1
.forEach(l -> add(sbjs, getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID)));
level2
.forEach(l -> add(sbjs, getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID)));
level3
.forEach(
l -> add(sbjs, getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID, true)));
level4
.forEach(
l -> add(sbjs, getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID, true)));
r.setSubject(sbjs);
r
.setDataInfo(
OafMapperUtils
.dataInfo(
false, null, true,
false,
OafMapperUtils
.qualifier(
ModelConstants.PROVENANCE_ENRICH,
null,
ModelConstants.DNET_PROVENANCE_ACTIONS,
ModelConstants.DNET_PROVENANCE_ACTIONS),
null));
return r;
}
private static void distributeFOSdois(SparkSession spark, String sourcePath, String outputPath) {
Dataset<FOSDataModel> fosDataset = readPath(spark, sourcePath, FOSDataModel.class);
fosDataset
.groupByKey((MapFunction<FOSDataModel, String>) v -> v.getDoi().toLowerCase(), Encoders.STRING())
.mapGroups((MapGroupsFunction<String, FOSDataModel, Result>) (k, it) -> {
Result r = new Result();
FOSDataModel first = it.next();
r.setId(DHPUtils.generateUnresolvedIdentifier(k, DOI));
HashSet<String> level1 = new HashSet<>();
HashSet<String> level2 = new HashSet<>();
HashSet<String> level3 = new HashSet<>();
HashSet<String> level4 = new HashSet<>();
addLevels(level1, level2, level3, level4, first);
it.forEachRemaining(v -> addLevels(level1, level2, level3, level4, v));
List<Subject> sbjs = new ArrayList<>();
level1
.forEach(l -> add(sbjs, getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID)));
level2
.forEach(l -> add(sbjs, getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID)));
level3
.forEach(
l -> add(sbjs, getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID, true)));
level4
.forEach(
l -> add(sbjs, getSubject(l, FOS_CLASS_ID, FOS_CLASS_NAME, UPDATE_SUBJECT_FOS_CLASS_ID, true)));
r.setSubject(sbjs);
r
.setDataInfo(
OafMapperUtils
.dataInfo(
false, null, true,
false,
OafMapperUtils
.qualifier(
ModelConstants.PROVENANCE_ENRICH,
null,
ModelConstants.DNET_PROVENANCE_ACTIONS,
ModelConstants.DNET_PROVENANCE_ACTIONS),
null));
return r;
}, Encoders.bean(Result.class))
.mapGroups(
(MapGroupsFunction<String, FOSDataModel, Result>) (k,
it) -> getResult(DHPUtils.generateUnresolvedIdentifier(k, DOI), it),
Encoders.bean(Result.class))
.write()
.mode(SaveMode.Overwrite)
.option("compression", "gzip")

View File

@ -0,0 +1,92 @@
package eu.dnetlib.dhp.actionmanager.fosnodoi;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession;
import java.io.IOException;
import java.io.Serializable;
import java.util.*;
import org.apache.commons.cli.ParseException;
import org.apache.commons.io.IOUtils;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.io.compress.GzipCodec;
import org.apache.hadoop.mapred.SequenceFileOutputFormat;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaPairRDD;
import org.apache.spark.api.java.function.FilterFunction;
import org.apache.spark.api.java.function.FlatMapFunction;
import org.apache.spark.api.java.function.MapFunction;
import org.apache.spark.sql.Encoders;
import org.apache.spark.sql.SparkSession;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.opencitations.model.COCI;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.schema.action.AtomicAction;
import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.oaf.Relation;
import eu.dnetlib.dhp.schema.oaf.Result;
import eu.dnetlib.dhp.schema.oaf.utils.*;
import scala.Tuple2;
public class CreateActionSetSparkJob implements Serializable {
private static final Logger log = LoggerFactory.getLogger(CreateActionSetSparkJob.class);
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
public static void main(final String[] args) throws IOException, ParseException {
final ArgumentApplicationParser parser = new ArgumentApplicationParser(
IOUtils
.toString(
Objects
.requireNonNull(
CreateActionSetSparkJob.class
.getResourceAsStream(
"/eu/dnetlib/dhp/actionmanager/fosnodoi/as_parameters.json"))));
parser.parseArgument(args);
Boolean isSparkSessionManaged = Optional
.ofNullable(parser.get("isSparkSessionManaged"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
final String inputPath = parser.get("sourcePath");
log.info("inputPath {}", inputPath);
final String outputPath = parser.get("outputPath");
log.info("outputPath {}", outputPath);
SparkConf conf = new SparkConf();
runWithSparkSession(
conf,
isSparkSessionManaged,
spark -> createActionSet(spark, inputPath, outputPath));
}
private static void createActionSet(SparkSession spark, String inputPath, String outputPath) {
spark
.read()
.textFile(inputPath)
.map(
(MapFunction<String, Result>) value -> OBJECT_MAPPER.readValue(value, Result.class),
Encoders.bean(Result.class))
.toJavaRDD()
.map(p -> new AtomicAction(p.getClass(), p))
.mapToPair(
aa -> new Tuple2<>(new Text(aa.getClazz().getCanonicalName()),
new Text(OBJECT_MAPPER.writeValueAsString(aa))))
.saveAsHadoopFile(
outputPath, Text.class, Text.class, SequenceFileOutputFormat.class, GzipCodec.class);
}
}

View File

@ -22,12 +22,14 @@ import org.apache.spark.sql.SparkSession;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.core.JsonProcessingException;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.opencitations.model.COCI;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.schema.action.AtomicAction;
import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.common.ModelSupport;
import eu.dnetlib.dhp.schema.oaf.*;
import eu.dnetlib.dhp.schema.oaf.utils.*;
import eu.dnetlib.dhp.utils.DHPUtils;
@ -37,16 +39,12 @@ public class CreateActionSetSparkJob implements Serializable {
public static final String OPENCITATIONS_CLASSID = "sysimport:crosswalk:opencitations";
public static final String OPENCITATIONS_CLASSNAME = "Imported from OpenCitations";
// DOI-to-DOI citations
public static final String COCI = "COCI";
// PMID-to-PMID citations
public static final String POCI = "POCI";
private static final String DOI_PREFIX = "50|doi_________::";
private static final String PMID_PREFIX = "50|pmid________::";
private static final String ARXIV_PREFIX = "50|arXiv_______::";
private static final String PMCID_PREFIX = "50|pmcid_______::";
private static final String TRUST = "0.91";
private static final Logger log = LoggerFactory.getLogger(CreateActionSetSparkJob.class);
@ -79,38 +77,30 @@ public class CreateActionSetSparkJob implements Serializable {
final String outputPath = parser.get("outputPath");
log.info("outputPath {}", outputPath);
final boolean shouldDuplicateRels = Optional
.ofNullable(parser.get("shouldDuplicateRels"))
.map(Boolean::valueOf)
.orElse(Boolean.FALSE);
SparkConf conf = new SparkConf();
runWithSparkSession(
conf,
isSparkSessionManaged,
spark -> extractContent(spark, inputPath, outputPath, shouldDuplicateRels));
spark -> extractContent(spark, inputPath, outputPath));
}
private static void extractContent(SparkSession spark, String inputPath, String outputPath,
boolean shouldDuplicateRels) {
private static void extractContent(SparkSession spark, String inputPath, String outputPath) {
getTextTextJavaPairRDD(spark, inputPath, shouldDuplicateRels, COCI)
.union(getTextTextJavaPairRDD(spark, inputPath, shouldDuplicateRels, POCI))
.saveAsHadoopFile(outputPath, Text.class, Text.class, SequenceFileOutputFormat.class, GzipCodec.class);
getTextTextJavaPairRDD(spark, inputPath)
.saveAsHadoopFile(outputPath, Text.class, Text.class, SequenceFileOutputFormat.class);// , GzipCodec.class);
}
private static JavaPairRDD<Text, Text> getTextTextJavaPairRDD(SparkSession spark, String inputPath,
boolean shouldDuplicateRels, String prefix) {
private static JavaPairRDD<Text, Text> getTextTextJavaPairRDD(SparkSession spark, String inputPath) {
return spark
.read()
.textFile(inputPath + "/" + prefix + "/" + prefix + "_JSON/*")
.textFile(inputPath)
.map(
(MapFunction<String, COCI>) value -> OBJECT_MAPPER.readValue(value, COCI.class),
Encoders.bean(COCI.class))
.flatMap(
(FlatMapFunction<COCI, Relation>) value -> createRelation(
value, shouldDuplicateRels, prefix)
value)
.iterator(),
Encoders.bean(Relation.class))
.filter((FilterFunction<Relation>) Objects::nonNull)
@ -121,34 +111,68 @@ public class CreateActionSetSparkJob implements Serializable {
new Text(OBJECT_MAPPER.writeValueAsString(aa))));
}
private static List<Relation> createRelation(COCI value, boolean duplicate, String p) {
private static List<Relation> createRelation(COCI value) throws JsonProcessingException {
List<Relation> relationList = new ArrayList<>();
String prefix;
String citing;
String cited;
switch (p) {
case COCI:
prefix = DOI_PREFIX;
citing = prefix
switch (value.getCiting_pid()) {
case "doi":
citing = DOI_PREFIX
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.doi.toString(), value.getCiting()));
cited = prefix
break;
case "pmid":
citing = PMID_PREFIX
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.pmid.toString(), value.getCiting()));
break;
case "arxiv":
citing = ARXIV_PREFIX
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.arXiv.toString(), value.getCiting()));
break;
case "pmcid":
citing = PMCID_PREFIX
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.pmc.toString(), value.getCiting()));
break;
case "isbn":
case "issn":
return relationList;
default:
throw new IllegalStateException("Invalid prefix: " + new ObjectMapper().writeValueAsString(value));
}
switch (value.getCited_pid()) {
case "doi":
cited = DOI_PREFIX
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.doi.toString(), value.getCited()));
break;
case POCI:
prefix = PMID_PREFIX;
citing = prefix
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.pmid.toString(), value.getCiting()));
cited = prefix
case "pmid":
cited = PMID_PREFIX
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.pmid.toString(), value.getCited()));
break;
case "arxiv":
cited = ARXIV_PREFIX
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.arXiv.toString(), value.getCited()));
break;
case "pmcid":
cited = PMCID_PREFIX
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.pmc.toString(), value.getCited()));
break;
case "isbn":
case "issn":
return relationList;
default:
throw new IllegalStateException("Invalid prefix: " + p);
throw new IllegalStateException("Invalid prefix: " + new ObjectMapper().writeValueAsString(value));
}
if (!citing.equals(cited)) {
@ -157,15 +181,6 @@ public class CreateActionSetSparkJob implements Serializable {
getRelation(
citing,
cited, ModelConstants.CITES));
if (duplicate && value.getCiting().endsWith(".refs")) {
citing = prefix + IdentifierFactory
.md5(
CleaningFunctions
.normalizePidValue(
"doi", value.getCiting().substring(0, value.getCiting().indexOf(".refs"))));
relationList.add(getRelation(citing, cited, ModelConstants.CITES));
}
}
return relationList;

View File

@ -12,10 +12,7 @@ import java.util.zip.ZipInputStream;
import org.apache.commons.cli.ParseException;
import org.apache.commons.io.IOUtils;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FSDataInputStream;
import org.apache.hadoop.fs.FSDataOutputStream;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.fs.*;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
@ -37,17 +34,17 @@ public class GetOpenCitationsRefs implements Serializable {
parser.parseArgument(args);
final String[] inputFile = parser.get("inputFile").split(";");
log.info("inputFile {}", Arrays.asList(inputFile));
// final String[] inputFile = parser.get("inputFile").split(";");
// log.info("inputFile {}", Arrays.asList(inputFile));
final String workingPath = parser.get("workingPath");
log.info("workingPath {}", workingPath);
final String inputPath = parser.get("inputPath");
log.info("inputPath {}", inputPath);
final String hdfsNameNode = parser.get("hdfsNameNode");
log.info("hdfsNameNode {}", hdfsNameNode);
final String prefix = parser.get("prefix");
log.info("prefix {}", prefix);
final String outputPath = parser.get("outputPath");
log.info("outputPath {}", outputPath);
Configuration conf = new Configuration();
conf.set("fs.defaultFS", hdfsNameNode);
@ -56,41 +53,42 @@ public class GetOpenCitationsRefs implements Serializable {
GetOpenCitationsRefs ocr = new GetOpenCitationsRefs();
for (String file : inputFile) {
ocr.doExtract(workingPath + "/Original/" + file, workingPath, fileSystem, prefix);
}
ocr.doExtract(inputPath, outputPath, fileSystem);
}
private void doExtract(String inputFile, String workingPath, FileSystem fileSystem, String prefix)
private void doExtract(String inputPath, String outputPath, FileSystem fileSystem)
throws IOException {
final Path path = new Path(inputFile);
RemoteIterator<LocatedFileStatus> fileStatusListIterator = fileSystem
.listFiles(
new Path(inputPath), true);
while (fileStatusListIterator.hasNext()) {
LocatedFileStatus fileStatus = fileStatusListIterator.next();
// do stuff with the file like ...
FSDataInputStream oc_zip = fileSystem.open(fileStatus.getPath());
try (ZipInputStream zis = new ZipInputStream(oc_zip)) {
ZipEntry entry = null;
while ((entry = zis.getNextEntry()) != null) {
FSDataInputStream oc_zip = fileSystem.open(path);
if (!entry.isDirectory()) {
String fileName = entry.getName();
// fileName = fileName.substring(0, fileName.indexOf("T")) + "_" + count;
fileName = fileName.substring(0, fileName.lastIndexOf("."));
// count++;
try (
FSDataOutputStream out = fileSystem
.create(new Path(outputPath + "/" + fileName + ".gz"));
GZIPOutputStream gzipOs = new GZIPOutputStream(new BufferedOutputStream(out))) {
// int count = 1;
try (ZipInputStream zis = new ZipInputStream(oc_zip)) {
ZipEntry entry = null;
while ((entry = zis.getNextEntry()) != null) {
if (!entry.isDirectory()) {
String fileName = entry.getName();
// fileName = fileName.substring(0, fileName.indexOf("T")) + "_" + count;
fileName = fileName.substring(0, fileName.lastIndexOf("."));
// count++;
try (
FSDataOutputStream out = fileSystem
.create(new Path(workingPath + "/" + prefix + "/" + fileName + ".gz"));
GZIPOutputStream gzipOs = new GZIPOutputStream(new BufferedOutputStream(out))) {
IOUtils.copy(zis, gzipOs);
IOUtils.copy(zis, gzipOs);
}
}
}
}
}
}

View File

@ -0,0 +1,171 @@
package eu.dnetlib.dhp.actionmanager.opencitations;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession;
import java.io.IOException;
import java.io.Serializable;
import java.util.Arrays;
import java.util.Objects;
import java.util.Optional;
import java.util.stream.Collectors;
import java.util.zip.ZipEntry;
import java.util.zip.ZipInputStream;
import org.apache.commons.cli.ParseException;
import org.apache.commons.io.IOUtils;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FSDataInputStream;
import org.apache.hadoop.fs.FSDataOutputStream;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.function.FlatMapFunction;
import org.apache.spark.api.java.function.ForeachFunction;
import org.apache.spark.api.java.function.MapFunction;
import org.apache.spark.sql.*;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.opencitations.model.COCI;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import scala.Tuple2;
/**
* @author miriam.baglioni
* @Date 29/02/24
*/
public class MapOCIdsInPids implements Serializable {
private static final Logger log = LoggerFactory.getLogger(CreateActionSetSparkJob.class);
private static final String DELIMITER = ",";
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
public static void main(final String[] args) throws IOException, ParseException {
final ArgumentApplicationParser parser = new ArgumentApplicationParser(
IOUtils
.toString(
Objects
.requireNonNull(
MapOCIdsInPids.class
.getResourceAsStream(
"/eu/dnetlib/dhp/actionmanager/opencitations/remap_parameters.json"))));
parser.parseArgument(args);
Boolean isSparkSessionManaged = Optional
.ofNullable(parser.get("isSparkSessionManaged"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
final String inputPath = parser.get("inputPath");
log.info("inputPath {}", inputPath);
final String outputPath = parser.get("outputPath");
log.info("outputPath {}", outputPath);
final String nameNode = parser.get("nameNode");
log.info("nameNode {}", nameNode);
unzipCorrespondenceFile(inputPath, nameNode);
SparkConf conf = new SparkConf();
runWithSparkSession(
conf,
isSparkSessionManaged,
spark -> mapIdentifiers(spark, inputPath, outputPath));
}
private static void unzipCorrespondenceFile(String inputPath, String hdfsNameNode) throws IOException {
Configuration conf = new Configuration();
conf.set("fs.defaultFS", hdfsNameNode);
final Path path = new Path(inputPath + "/correspondence/omid.zip");
FileSystem fileSystem = FileSystem.get(conf);
FSDataInputStream project_zip = fileSystem.open(path);
try (ZipInputStream zis = new ZipInputStream(project_zip)) {
ZipEntry entry = null;
while ((entry = zis.getNextEntry()) != null) {
if (!entry.isDirectory()) {
String fileName = entry.getName();
byte buffer[] = new byte[1024];
int count;
try (
FSDataOutputStream out = fileSystem
.create(new Path(inputPath + "/correspondence/omid.csv"))) {
while ((count = zis.read(buffer, 0, buffer.length)) != -1)
out.write(buffer, 0, count);
}
}
}
}
}
private static void mapIdentifiers(SparkSession spark, String inputPath, String outputPath) {
Dataset<COCI> coci = spark
.read()
.textFile(inputPath + "/JSON")
.map(
(MapFunction<String, COCI>) value -> OBJECT_MAPPER.readValue(value, COCI.class),
Encoders.bean(COCI.class));
Dataset<Tuple2<String, String>> correspondenceData = spark
.read()
.format("csv")
.option("sep", DELIMITER)
.option("inferSchema", "true")
.option("header", "true")
.option("quotes", "\"")
.load(inputPath + "/correspondence/omid.csv")
.repartition(5000)
.flatMap((FlatMapFunction<Row, Tuple2<String, String>>) r -> {
String ocIdentifier = r.getAs("omid");
String[] correspondentIdentifiers = ((String) r.getAs("id")).split(" ");
return Arrays
.stream(correspondentIdentifiers)
.map(ci -> new Tuple2<String, String>(ocIdentifier, ci))
.collect(Collectors.toList())
.iterator();
}, Encoders.tuple(Encoders.STRING(), Encoders.STRING()));
Dataset<COCI> mappedCitingDataset = coci
.joinWith(correspondenceData, coci.col("citing").equalTo(correspondenceData.col("_1")))
.map((MapFunction<Tuple2<COCI, Tuple2<String, String>>, COCI>) t2 -> {
String correspondent = t2._2()._2();
t2._1().setCiting_pid(correspondent.substring(0, correspondent.indexOf(":")));
t2._1().setCiting(correspondent.substring(correspondent.indexOf(":") + 1));
return t2._1();
}, Encoders.bean(COCI.class));
mappedCitingDataset
.joinWith(correspondenceData, mappedCitingDataset.col("cited").equalTo(correspondenceData.col("_1")))
.map((MapFunction<Tuple2<COCI, Tuple2<String, String>>, COCI>) t2 -> {
String correspondent = t2._2()._2();
t2._1().setCited_pid(correspondent.substring(0, correspondent.indexOf(":")));
t2._1().setCited(correspondent.substring(correspondent.indexOf(":") + 1));
return t2._1();
}, Encoders.bean(COCI.class))
.write()
.mode(SaveMode.Append)
.option("compression", "gzip")
.json(outputPath);
}
}

View File

@ -12,11 +12,9 @@ import java.util.Optional;
import org.apache.commons.io.IOUtils;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.LocatedFileStatus;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.fs.RemoteIterator;
import org.apache.hadoop.fs.*;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.function.FilterFunction;
import org.apache.spark.api.java.function.MapFunction;
import org.apache.spark.sql.*;
import org.slf4j.Logger;
@ -42,19 +40,21 @@ public class ReadCOCI implements Serializable {
final String outputPath = parser.get("outputPath");
log.info("outputPath: {}", outputPath);
final String[] inputFile = parser.get("inputFile").split(";");
log.info("inputFile {}", Arrays.asList(inputFile));
final String hdfsNameNode = parser.get("hdfsNameNode");
log.info("hdfsNameNode {}", hdfsNameNode);
Boolean isSparkSessionManaged = isSparkSessionManaged(parser);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
final String workingPath = parser.get("workingPath");
final String workingPath = parser.get("inputPath");
log.info("workingPath {}", workingPath);
final String format = parser.get("format");
log.info("format {}", format);
SparkConf sconf = new SparkConf();
Configuration conf = new Configuration();
conf.set("fs.defaultFS", hdfsNameNode);
FileSystem fileSystem = FileSystem.get(conf);
final String delimiter = Optional
.ofNullable(parser.get("delimiter"))
.orElse(DEFAULT_DELIMITER);
@ -66,20 +66,21 @@ public class ReadCOCI implements Serializable {
doRead(
spark,
workingPath,
inputFile,
fileSystem,
outputPath,
delimiter,
format);
delimiter);
});
}
private static void doRead(SparkSession spark, String workingPath, String[] inputFiles,
private static void doRead(SparkSession spark, String workingPath, FileSystem fileSystem,
String outputPath,
String delimiter, String format) {
for (String inputFile : inputFiles) {
String pString = workingPath + "/" + inputFile + ".gz";
String delimiter) throws IOException {
RemoteIterator<LocatedFileStatus> fileStatusListIterator = fileSystem
.listFiles(
new Path(workingPath), true);
while (fileStatusListIterator.hasNext()) {
LocatedFileStatus fileStatus = fileStatusListIterator.next();
log.info("extracting file {}", fileStatus.getPath().toString());
Dataset<Row> cociData = spark
.read()
.format("csv")
@ -87,26 +88,26 @@ public class ReadCOCI implements Serializable {
.option("inferSchema", "true")
.option("header", "true")
.option("quotes", "\"")
.load(pString)
.load(fileStatus.getPath().toString())
.repartition(100);
cociData.map((MapFunction<Row, COCI>) row -> {
COCI coci = new COCI();
if (format.equals("COCI")) {
coci.setCiting(row.getString(1));
coci.setCited(row.getString(2));
} else {
coci.setCiting(String.valueOf(row.getInt(1)));
coci.setCited(String.valueOf(row.getInt(2)));
}
coci.setCiting(row.getString(1));
coci.setCited(row.getString(2));
coci.setOci(row.getString(0));
return coci;
}, Encoders.bean(COCI.class))
.filter((FilterFunction<COCI>) c -> c != null)
.write()
.mode(SaveMode.Overwrite)
.mode(SaveMode.Append)
.option("compression", "gzip")
.json(outputPath + inputFile);
.json(outputPath);
fileSystem.rename(fileStatus.getPath(), new Path("/tmp/miriam/OC/DONE"));
}
}

View File

@ -9,8 +9,10 @@ public class COCI implements Serializable {
private String oci;
private String citing;
private String citing_pid;
private String cited;
private String cited_pid;
public String getOci() {
return oci;
@ -25,6 +27,8 @@ public class COCI implements Serializable {
}
public void setCiting(String citing) {
if (citing != null && citing.startsWith("omid:"))
citing = citing.substring(5);
this.citing = citing;
}
@ -33,7 +37,24 @@ public class COCI implements Serializable {
}
public void setCited(String cited) {
if (cited != null && cited.startsWith("omid:"))
cited = cited.substring(5);
this.cited = cited;
}
public String getCiting_pid() {
return citing_pid;
}
public void setCiting_pid(String citing_pid) {
this.citing_pid = citing_pid;
}
public String getCited_pid() {
return cited_pid;
}
public void setCited_pid(String cited_pid) {
this.cited_pid = cited_pid;
}
}

View File

@ -1,12 +1,20 @@
package eu.dnetlib.dhp.actionmanager.project;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession;
import java.util.Arrays;
import java.util.Objects;
import java.util.Optional;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.project.utils.model.CSVProgramme;
import eu.dnetlib.dhp.actionmanager.project.utils.model.CSVProject;
import eu.dnetlib.dhp.actionmanager.project.utils.model.JsonTopic;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.common.HdfsSupport;
import eu.dnetlib.dhp.schema.action.AtomicAction;
import eu.dnetlib.dhp.schema.common.ModelSupport;
import eu.dnetlib.dhp.schema.oaf.H2020Classification;
import eu.dnetlib.dhp.schema.oaf.H2020Programme;
import eu.dnetlib.dhp.schema.oaf.OafEntity;
import eu.dnetlib.dhp.schema.oaf.Project;
import eu.dnetlib.dhp.schema.oaf.utils.MergeUtils;
import eu.dnetlib.dhp.utils.DHPUtils;
import org.apache.commons.io.IOUtils;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapred.SequenceFileOutputFormat;
@ -18,24 +26,14 @@ import org.apache.spark.sql.Encoders;
import org.apache.spark.sql.SparkSession;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.project.utils.model.CSVProgramme;
import eu.dnetlib.dhp.actionmanager.project.utils.model.CSVProject;
import eu.dnetlib.dhp.actionmanager.project.utils.model.EXCELTopic;
import eu.dnetlib.dhp.actionmanager.project.utils.model.JsonTopic;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.common.HdfsSupport;
import eu.dnetlib.dhp.schema.action.AtomicAction;
import eu.dnetlib.dhp.schema.common.ModelSupport;
import eu.dnetlib.dhp.schema.oaf.H2020Classification;
import eu.dnetlib.dhp.schema.oaf.H2020Programme;
import eu.dnetlib.dhp.schema.oaf.OafEntity;
import eu.dnetlib.dhp.schema.oaf.Project;
import eu.dnetlib.dhp.utils.DHPUtils;
import scala.Tuple2;
import java.util.Arrays;
import java.util.Objects;
import java.util.Optional;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession;
/**
* Class that makes the ActionSet. To prepare the AS two joins are needed
*
@ -160,9 +158,11 @@ public class SparkAtomicActionJob {
(MapFunction<Project, String>) OafEntity::getId,
Encoders.STRING())
.mapGroups((MapGroupsFunction<String, Project, Project>) (s, it) -> {
Project first = it.next();
it.forEachRemaining(first::mergeFrom);
return first;
Project merge = it.next();
while (it.hasNext()) {
merge = MergeUtils.mergeProject(merge, it.next());
}
return merge;
}, Encoders.bean(Project.class))
.toJavaRDD()
.map(p -> new AtomicAction(Project.class, p))

View File

@ -0,0 +1,195 @@
package eu.dnetlib.dhp.actionmanager.transformativeagreement;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession;
import java.io.IOException;
import java.io.Serializable;
import java.util.*;
import org.apache.commons.cli.ParseException;
import org.apache.commons.io.IOUtils;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.io.compress.GzipCodec;
import org.apache.hadoop.mapred.SequenceFileOutputFormat;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.function.FilterFunction;
import org.apache.spark.api.java.function.FlatMapFunction;
import org.apache.spark.api.java.function.MapFunction;
import org.apache.spark.sql.Encoders;
import org.apache.spark.sql.SparkSession;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.transformativeagreement.model.TransformativeAgreementModel;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.schema.action.AtomicAction;
import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.oaf.Country;
import eu.dnetlib.dhp.schema.oaf.Relation;
import eu.dnetlib.dhp.schema.oaf.Result;
import eu.dnetlib.dhp.schema.oaf.utils.*;
import scala.Tuple2;
public class CreateActionSetSparkJob implements Serializable {
private static final Logger log = LoggerFactory.getLogger(CreateActionSetSparkJob.class);
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
private static final String IREL_PROJECT = "40|100018998___::1e5e62235d094afd01cd56e65112fc63";
private static final String TRANSFORMATIVE_AGREEMENT = "openapc::transformativeagreement";
public static void main(final String[] args) throws IOException, ParseException {
final ArgumentApplicationParser parser = new ArgumentApplicationParser(
IOUtils
.toString(
Objects
.requireNonNull(
CreateActionSetSparkJob.class
.getResourceAsStream(
"/eu/dnetlib/dhp/actionmanager/transformativeagreement/as_parameters.json"))));
parser.parseArgument(args);
Boolean isSparkSessionManaged = Optional
.ofNullable(parser.get("isSparkSessionManaged"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
final String inputPath = parser.get("inputPath");
log.info("inputPath {}", inputPath);
final String outputPath = parser.get("outputPath");
log.info("outputPath {}", outputPath);
SparkConf conf = new SparkConf();
runWithSparkSession(
conf,
isSparkSessionManaged,
spark -> createActionSet(spark, inputPath, outputPath));
}
private static void createActionSet(SparkSession spark, String inputPath, String outputPath) {
JavaRDD<AtomicAction> relations = spark
.read()
.textFile(inputPath)
.map(
(MapFunction<String, TransformativeAgreementModel>) value -> OBJECT_MAPPER
.readValue(value, TransformativeAgreementModel.class),
Encoders.bean(TransformativeAgreementModel.class))
.flatMap(
(FlatMapFunction<TransformativeAgreementModel, Relation>) value -> createRelation(
value)
.iterator(),
Encoders.bean(Relation.class))
.filter((FilterFunction<Relation>) Objects::nonNull)
.toJavaRDD()
.map(p -> new AtomicAction(p.getClass(), p));
//TODO relations in stand-by waiting to know if we need to create them or not In case we need just make a union before saving the sequence file
spark
.read()
.textFile(inputPath)
.map(
(MapFunction<String, TransformativeAgreementModel>) value -> OBJECT_MAPPER
.readValue(value, TransformativeAgreementModel.class),
Encoders.bean(TransformativeAgreementModel.class))
.map(
(MapFunction<TransformativeAgreementModel, Result>) value -> createResult(
value),
Encoders.bean(Result.class))
.filter((FilterFunction<Result>) r -> r != null)
.toJavaRDD()
.map(p -> new AtomicAction(p.getClass(), p))
.mapToPair(
aa -> new Tuple2<>(new Text(aa.getClazz().getCanonicalName()),
new Text(OBJECT_MAPPER.writeValueAsString(aa))))
.saveAsHadoopFile(
outputPath, Text.class, Text.class, SequenceFileOutputFormat.class, GzipCodec.class);
}
private static Result createResult(TransformativeAgreementModel value) {
Result r = new Result();
r
.setId(
"50|doi_________::"
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.doi.toString(), value.getDoi())));
r.setTransformativeAgreement(value.getAgreement());
Country country = new Country();
country.setClassid(value.getCountry());
country.setClassname(value.getCountry());
country
.setDataInfo(
OafMapperUtils
.dataInfo(
false, ModelConstants.SYSIMPORT_ACTIONSET, false, false,
OafMapperUtils
.qualifier(
"openapc::transformativeagreement",
"Harvested from Trnasformative Agreement file from OpenAPC",
ModelConstants.DNET_PROVENANCE_ACTIONS, ModelConstants.DNET_PROVENANCE_ACTIONS),
"0.9"));
country.setSchemeid(ModelConstants.DNET_COUNTRY_TYPE);
country.setSchemename(ModelConstants.DNET_COUNTRY_TYPE);
r.setCountry(Arrays.asList(country));
return r;
}
private static List<Relation> createRelation(TransformativeAgreementModel value) {
List<Relation> relationList = new ArrayList<>();
if (value.getAgreement().startsWith("IReL")) {
String paper;
paper = "50|doi_________::"
+ IdentifierFactory
.md5(PidCleaner.normalizePidValue(PidType.doi.toString(), value.getDoi()));
relationList
.add(
getRelation(
paper,
IREL_PROJECT, ModelConstants.IS_PRODUCED_BY));
relationList.add(getRelation(IREL_PROJECT, paper, ModelConstants.PRODUCES));
}
return relationList;
}
public static Relation getRelation(
String source,
String target,
String relClass) {
return OafMapperUtils
.getRelation(
source,
target,
ModelConstants.RESULT_PROJECT,
ModelConstants.OUTCOME,
relClass,
Arrays
.asList(
OafMapperUtils.keyValue(ModelConstants.OPEN_APC_ID, ModelConstants.OPEN_APC_NAME)),
OafMapperUtils
.dataInfo(
false, null, false, false,
OafMapperUtils
.qualifier(
TRANSFORMATIVE_AGREEMENT, "Transformative Agreement",
ModelConstants.DNET_PROVENANCE_ACTIONS, ModelConstants.DNET_PROVENANCE_ACTIONS),
"0.9"),
null);
}
}

View File

@ -0,0 +1,51 @@
package eu.dnetlib.dhp.actionmanager.transformativeagreement.model;
import java.io.Serializable;
import com.fasterxml.jackson.annotation.JsonIgnoreProperties;
/**
* @author miriam.baglioni
* @Date 18/12/23
*/
@JsonIgnoreProperties(ignoreUnknown = true)
public class TransformativeAgreementModel implements Serializable {
private String institution;
private String doi;
private String agreement;
private String country;
public String getCountry() {
return country;
}
public void setCountry(String country) {
this.country = country;
}
public String getInstitution() {
return institution;
}
public void setInstitution(String institution) {
this.institution = institution;
}
public String getDoi() {
return doi;
}
public void setDoi(String doi) {
this.doi = doi;
}
public String getAgreement() {
return agreement;
}
public void setAgreement(String agreement) {
this.agreement = agreement;
}
}

View File

@ -19,6 +19,7 @@ import org.slf4j.LoggerFactory;
import eu.dnetlib.dhp.aggregation.common.ReporterCallback;
import eu.dnetlib.dhp.aggregation.common.ReportingJob;
import eu.dnetlib.dhp.collection.plugin.CollectorPlugin;
import eu.dnetlib.dhp.collection.plugin.base.BaseCollectorPlugin;
import eu.dnetlib.dhp.collection.plugin.file.FileCollectorPlugin;
import eu.dnetlib.dhp.collection.plugin.file.FileGZipCollectorPlugin;
import eu.dnetlib.dhp.collection.plugin.mongodb.MDStoreCollectorPlugin;
@ -120,6 +121,8 @@ public class CollectorWorker extends ReportingJob {
return new FileCollectorPlugin(fileSystem);
case fileGzip:
return new FileGZipCollectorPlugin(fileSystem);
case baseDump:
return new BaseCollectorPlugin(this.fileSystem);
case other:
final CollectorPlugin.NAME.OTHER_NAME plugin = Optional
.ofNullable(api.getParams().get("other_plugin_type"))

View File

@ -10,7 +10,8 @@ import eu.dnetlib.dhp.common.collection.CollectorException;
public interface CollectorPlugin {
enum NAME {
oai, other, rest_json2xml, file, fileGzip;
oai, other, rest_json2xml, file, fileGzip, baseDump;
public enum OTHER_NAME {
mdstore_mongodb_dump, mdstore_mongodb

View File

@ -0,0 +1,171 @@
package eu.dnetlib.dhp.collection.plugin.base;
import java.io.BufferedInputStream;
import java.io.ByteArrayInputStream;
import java.io.InputStream;
import java.io.StringWriter;
import java.util.Iterator;
import java.util.concurrent.BlockingQueue;
import java.util.concurrent.LinkedBlockingQueue;
import javax.xml.stream.XMLEventReader;
import javax.xml.stream.XMLEventWriter;
import javax.xml.stream.XMLInputFactory;
import javax.xml.stream.XMLOutputFactory;
import javax.xml.stream.events.EndElement;
import javax.xml.stream.events.StartElement;
import javax.xml.stream.events.XMLEvent;
import org.apache.commons.compress.archivers.tar.TarArchiveEntry;
import org.apache.commons.compress.archivers.tar.TarArchiveInputStream;
import org.apache.commons.compress.compressors.CompressorInputStream;
import org.apache.commons.compress.compressors.CompressorStreamFactory;
import org.apache.commons.io.IOUtils;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import eu.dnetlib.dhp.common.aggregation.AggregatorReport;
public class BaseCollectorIterator implements Iterator<String> {
private String nextElement;
private final BlockingQueue<String> queue = new LinkedBlockingQueue<>(100);
private static final Logger log = LoggerFactory.getLogger(BaseCollectorIterator.class);
private static final String END_ELEM = "__END__";
public BaseCollectorIterator(final FileSystem fs, final Path filePath, final AggregatorReport report) {
new Thread(() -> importHadoopFile(fs, filePath, report)).start();
try {
this.nextElement = this.queue.take();
} catch (final InterruptedException e) {
throw new RuntimeException(e);
}
}
protected BaseCollectorIterator(final String resourcePath, final AggregatorReport report) {
new Thread(() -> importTestFile(resourcePath, report)).start();
try {
this.nextElement = this.queue.take();
} catch (final InterruptedException e) {
throw new RuntimeException(e);
}
}
@Override
public synchronized boolean hasNext() {
return (this.nextElement != null) & !END_ELEM.equals(this.nextElement);
}
@Override
public synchronized String next() {
try {
return END_ELEM.equals(this.nextElement) ? null : this.nextElement;
} finally {
try {
this.nextElement = this.queue.take();
} catch (final InterruptedException e) {
throw new RuntimeException(e);
}
}
}
private void importHadoopFile(final FileSystem fs, final Path filePath, final AggregatorReport report) {
log.info("I start to read the TAR stream");
try (InputStream origInputStream = fs.open(filePath);
final TarArchiveInputStream tarInputStream = new TarArchiveInputStream(origInputStream)) {
importTarStream(tarInputStream, report);
} catch (final Throwable e) {
throw new RuntimeException("Error processing BASE records", e);
}
}
private void importTestFile(final String resourcePath, final AggregatorReport report) {
try (final InputStream origInputStream = BaseCollectorIterator.class.getResourceAsStream(resourcePath);
final TarArchiveInputStream tarInputStream = new TarArchiveInputStream(origInputStream)) {
importTarStream(tarInputStream, report);
} catch (final Throwable e) {
throw new RuntimeException("Error processing BASE records", e);
}
}
private void importTarStream(final TarArchiveInputStream tarInputStream, final AggregatorReport report) {
long count = 0;
final XMLInputFactory xmlInputFactory = XMLInputFactory.newInstance();
final XMLOutputFactory xmlOutputFactory = XMLOutputFactory.newInstance();
try {
TarArchiveEntry entry;
while ((entry = (TarArchiveEntry) tarInputStream.getNextEntry()) != null) {
final String name = entry.getName();
if (!entry.isDirectory() && name.contains("ListRecords") && name.endsWith(".bz2")) {
log.info("Processing file (BZIP): " + name);
final byte[] bzipData = new byte[(int) entry.getSize()];
IOUtils.readFully(tarInputStream, bzipData);
try (InputStream bzipIs = new ByteArrayInputStream(bzipData);
final BufferedInputStream bzipBis = new BufferedInputStream(bzipIs);
final CompressorInputStream bzipInput = new CompressorStreamFactory()
.createCompressorInputStream(bzipBis)) {
final XMLEventReader reader = xmlInputFactory.createXMLEventReader(bzipInput);
XMLEventWriter eventWriter = null;
StringWriter xmlWriter = null;
while (reader.hasNext()) {
final XMLEvent nextEvent = reader.nextEvent();
if (nextEvent.isStartElement()) {
final StartElement startElement = nextEvent.asStartElement();
if ("record".equals(startElement.getName().getLocalPart())) {
xmlWriter = new StringWriter();
eventWriter = xmlOutputFactory.createXMLEventWriter(xmlWriter);
}
}
if (eventWriter != null) {
eventWriter.add(nextEvent);
}
if (nextEvent.isEndElement()) {
final EndElement endElement = nextEvent.asEndElement();
if ("record".equals(endElement.getName().getLocalPart())) {
eventWriter.flush();
eventWriter.close();
this.queue.put(xmlWriter.toString());
eventWriter = null;
xmlWriter = null;
count++;
}
}
}
}
}
}
this.queue.put(END_ELEM); // TO INDICATE THE END OF THE QUEUE
} catch (final Throwable e) {
log.error("Error processing BASE records", e);
report.put(e.getClass().getName(), e.getMessage());
throw new RuntimeException("Error processing BASE records", e);
} finally {
log.info("Total records (written in queue): " + count);
}
}
}

View File

@ -0,0 +1,159 @@
package eu.dnetlib.dhp.collection.plugin.base;
import java.io.IOException;
import java.sql.SQLException;
import java.util.HashSet;
import java.util.Iterator;
import java.util.Optional;
import java.util.Set;
import java.util.Spliterator;
import java.util.Spliterators;
import java.util.stream.Stream;
import java.util.stream.StreamSupport;
import org.apache.commons.io.IOUtils;
import org.apache.commons.lang3.StringUtils;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.dom4j.Document;
import org.dom4j.DocumentException;
import org.dom4j.DocumentHelper;
import org.dom4j.Node;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import eu.dnetlib.dhp.collection.ApiDescriptor;
import eu.dnetlib.dhp.collection.plugin.CollectorPlugin;
import eu.dnetlib.dhp.collection.plugin.file.AbstractSplittedRecordPlugin;
import eu.dnetlib.dhp.common.DbClient;
import eu.dnetlib.dhp.common.aggregation.AggregatorReport;
import eu.dnetlib.dhp.common.collection.CollectorException;
public class BaseCollectorPlugin implements CollectorPlugin {
private final FileSystem fs;
private static final Logger log = LoggerFactory.getLogger(AbstractSplittedRecordPlugin.class);
// MAPPING AND FILTERING ARE DEFINED HERE:
// https://docs.google.com/document/d/1Aj-ZAV11b44MCrAAUCPiS2TUlXb6PnJEu1utCMAcCOU/edit
public BaseCollectorPlugin(final FileSystem fs) {
this.fs = fs;
}
@Override
public Stream<String> collect(final ApiDescriptor api, final AggregatorReport report) throws CollectorException {
// the path of the dump file on HDFS
// http://oai.base-search.net/initial_load/base_oaipmh_dump-current.tar
// it could be downloaded from iis-cdh5-test-gw.ocean.icm.edu.pl and then copied on HDFS
final Path filePath = Optional
.ofNullable(api.getBaseUrl())
.map(Path::new)
.orElseThrow(() -> new CollectorException("missing baseUrl"));
// get the parameters for the connection to the OpenAIRE database.
// the database is used to obtain the list of the datasources that the plugin will collect
final String dbUrl = api.getParams().get("dbUrl");
final String dbUser = api.getParams().get("dbUser");
final String dbPassword = api.getParams().get("dbPassword");
// the types(comma separated, empty value for all) that the plugin will collect,
// the types should be expressed in the format of the normalized types of BASE (for example 1,121,...)
final String acceptedNormTypesString = api.getParams().get("acceptedNormTypes");
log.info("baseUrl: {}", filePath);
log.info("dbUrl: {}", dbUrl);
log.info("dbUser: {}", dbUser);
log.info("dbPassword: {}", "***");
log.info("acceptedNormTypes: {}", acceptedNormTypesString);
try {
if (!this.fs.exists(filePath)) {
throw new CollectorException("path does not exist: " + filePath);
}
} catch (final Throwable e) {
throw new CollectorException(e);
}
final Set<String> acceptedOpendoarIds = findAcceptedOpendoarIds(dbUrl, dbUser, dbPassword);
final Set<String> acceptedNormTypes = new HashSet<>();
if (StringUtils.isNotBlank(acceptedNormTypesString)) {
for (final String s : StringUtils.split(acceptedNormTypesString, ",")) {
if (StringUtils.isNotBlank(s)) {
acceptedNormTypes.add(s.trim());
}
}
}
final Iterator<String> iterator = new BaseCollectorIterator(this.fs, filePath, report);
final Spliterator<String> spliterator = Spliterators.spliteratorUnknownSize(iterator, Spliterator.ORDERED);
return StreamSupport
.stream(spliterator, false)
.filter(doc -> filterXml(doc, acceptedOpendoarIds, acceptedNormTypes));
}
private Set<String> findAcceptedOpendoarIds(final String dbUrl, final String dbUser, final String dbPassword)
throws CollectorException {
final Set<String> accepted = new HashSet<>();
try (final DbClient dbClient = new DbClient(dbUrl, dbUser, dbPassword)) {
final String sql = IOUtils
.toString(
getClass().getResourceAsStream("/eu/dnetlib/dhp/collection/plugin/base/sql/opendoar-accepted.sql"));
dbClient.processResults(sql, row -> {
try {
final String dsId = row.getString("id");
log.info("Accepted Datasource: " + dsId);
accepted.add(dsId);
} catch (final SQLException e) {
log.error("Error in SQL", e);
throw new RuntimeException("Error in SQL", e);
}
});
} catch (final IOException e) {
log.error("Error accessong SQL", e);
throw new CollectorException("Error accessong SQL", e);
}
log.info("Accepted Datasources (TOTAL): " + accepted.size());
return accepted;
}
protected static boolean filterXml(final String xml,
final Set<String> acceptedOpendoarIds,
final Set<String> acceptedNormTypes) {
try {
final Document doc = DocumentHelper.parseText(xml);
final String id = doc.valueOf("//*[local-name()='collection']/@opendoar_id").trim();
if (StringUtils.isBlank(id) || !acceptedOpendoarIds.contains("opendoar____::" + id)) {
return false;
}
if (acceptedNormTypes.isEmpty()) {
return true;
}
for (final Object s : doc.selectNodes("//*[local-name()='typenorm']")) {
if (acceptedNormTypes.contains(((Node) s).getText().trim())) {
return true;
}
}
return false;
} catch (final DocumentException e) {
log.error("Error parsing document", e);
throw new RuntimeException("Error parsing document", e);
}
}
}

View File

@ -52,8 +52,6 @@ public class RestIterator implements Iterator<String> {
private final String BASIC = "basic";
private final JsonUtils jsonUtils;
private final String baseUrl;
private final String resumptionType;
private final String resumptionParam;
@ -106,7 +104,6 @@ public class RestIterator implements Iterator<String> {
final String resultOutputFormat) {
this.clientParams = clientParams;
this.jsonUtils = new JsonUtils();
this.baseUrl = baseUrl;
this.resumptionType = resumptionType;
this.resumptionParam = resumptionParam;
@ -126,6 +123,7 @@ public class RestIterator implements Iterator<String> {
} catch (Exception e) {
throw new IllegalStateException("xml transformation init failed: " + e.getMessage());
}
initQueue();
}
@ -190,7 +188,7 @@ public class RestIterator implements Iterator<String> {
String resultJson;
String resultXml = "<?xml version=\"1.0\" encoding=\"UTF-8\"?>";
String nextQuery = "";
String emptyXml = resultXml + "<" + JsonUtils.wrapName + "></" + JsonUtils.wrapName + ">";
String emptyXml = resultXml + "<" + JsonUtils.XML_WRAP_TAG + "></" + JsonUtils.XML_WRAP_TAG + ">";
Node resultNode = null;
NodeList nodeList = null;
String qUrlArgument = "";
@ -231,7 +229,7 @@ public class RestIterator implements Iterator<String> {
resultStream = theHttpInputStream;
if ("json".equals(resultOutputFormat)) {
resultJson = IOUtils.toString(resultStream, StandardCharsets.UTF_8);
resultXml = jsonUtils.convertToXML(resultJson);
resultXml = JsonUtils.convertToXML(resultJson);
resultStream = IOUtils.toInputStream(resultXml, UTF_8);
}

View File

@ -3,82 +3,142 @@ package eu.dnetlib.dhp.collection.plugin.utils;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.json.JSONArray;
import org.json.JSONObject;
public class JsonUtils {
public static final String XML_WRAP_TAG = "recordWrap";
private static final String XML_HEADER = "<?xml version=\"1.0\" encoding=\"UTF-8\"?>";
private static final String INVALID_XMLTAG_CHARS = "!\"#$%&'()*+,/;<=>?@[\\]^`{|}~,";
private static final Log log = LogFactory.getLog(JsonUtils.class);
public static final String wrapName = "recordWrap";
/**
* convert in JSON-KeyName 'whitespace(s)' to '_' and '/' to '_', '(' and ')' to ''
* cleanup in JSON-KeyName
* check W3C XML syntax: https://www.w3.org/TR/2006/REC-xml11-20060816/#sec-starttags for valid tag names
* and work-around for the JSON to XML converting of org.json.XML-package.
*
* known bugs: doesn't prevent "key name":" ["sexy name",": penari","erotic dance"],
*
* @param jsonInput
* @return convertedJsonKeynameOutput
* @param input
* @return converted json object
*/
public String syntaxConvertJsonKeyNames(String jsonInput) {
log.trace("before convertJsonKeyNames: " + jsonInput);
// pre-clean json - rid spaces of element names (misinterpreted as elements with attributes in xml)
// replace ' 's in JSON Namens with '_'
while (jsonInput.matches(".*\"([^\"]*)\\s+([^\"]*)\":.*")) {
jsonInput = jsonInput.replaceAll("\"([^\"]*)\\s+([^\"]*)\":", "\"$1_$2\":");
public static JSONObject cleanJsonObject(final JSONObject input) {
if (null == input) {
return null;
}
// replace forward-slash (sign '/' ) in JSON Names with '_'
while (jsonInput.matches(".*\"([^\"]*)/([^\"]*)\":.*")) {
jsonInput = jsonInput.replaceAll("\"([^\"]*)/([^\"]*)\":", "\"$1_$2\":");
JSONObject result = new JSONObject();
for (String key : input.keySet()) {
Object value = input.opt(key);
if (value != null) {
result.put(cleanKey(key), cleanValue(value));
}
}
// replace '(' in JSON Names with ''
while (jsonInput.matches(".*\"([^\"]*)[(]([^\"]*)\":.*")) {
jsonInput = jsonInput.replaceAll("\"([^\"]*)[(]([^\"]*)\":", "\"$1$2\":");
}
// replace ')' in JSON Names with ''
while (jsonInput.matches(".*\"([^\"]*)[)]([^\"]*)\":.*")) {
jsonInput = jsonInput.replaceAll("\"([^\"]*)[)]([^\"]*)\":", "\"$1$2\":");
}
// add prefix of startNumbers in JSON Keynames with 'n_'
while (jsonInput.matches(".*\"([^\"][0-9])([^\"]*)\":.*")) {
jsonInput = jsonInput.replaceAll("\"([^\"][0-9])([^\"]*)\":", "\"n_$1$2\":");
}
// add prefix of only numbers in JSON Keynames with 'm_'
while (jsonInput.matches(".*\"([0-9]+)\":.*")) {
jsonInput = jsonInput.replaceAll("\"([0-9]+)\":", "\"m_$1\":");
}
// replace ':' between number like '2018-08-28T11:05:00Z' in JSON keynames with ''
while (jsonInput.matches(".*\"([^\"]*[0-9]):([0-9][^\"]*)\":.*")) {
jsonInput = jsonInput.replaceAll("\"([^\"]*[0-9]):([0-9][^\"]*)\":", "\"$1$2\":");
}
// replace ',' in JSON Keynames with '.' to prevent , in xml tagnames.
// while (jsonInput.matches(".*\"([^\"]*),([^\"]*)\":.*")) {
// jsonInput = jsonInput.replaceAll("\"([^\"]*),([^\"]*)\":", "\"$1.$2\":");
// }
// replace '=' in JSON Keynames with '-'
while (jsonInput.matches(".*\"([^\"]*)=([^\"]*)\":.*")) {
jsonInput = jsonInput.replaceAll("\"([^\"]*)=([^\"]*)\":", "\"$1-$2\":");
}
log.trace("after syntaxConvertJsonKeyNames: " + jsonInput);
return jsonInput;
return result;
}
public String convertToXML(final String jsonRecord) {
String resultXml = "<?xml version=\"1.0\" encoding=\"UTF-8\"?>";
org.json.JSONObject jsonObject = new org.json.JSONObject(syntaxConvertJsonKeyNames(jsonRecord));
resultXml += org.json.XML.toString(jsonObject, wrapName); // wrap xml in single root element
log.trace("before inputStream: " + resultXml);
resultXml = XmlCleaner.cleanAllEntities(resultXml);
log.trace("after cleaning: " + resultXml);
return resultXml;
private static Object cleanValue(Object object) {
if (object instanceof JSONObject) {
return cleanJsonObject((JSONObject) object);
} else if (object instanceof JSONArray) {
JSONArray array = (JSONArray) object;
JSONArray res = new JSONArray();
for (int i = array.length() - 1; i >= 0; i--) {
res.put(i, cleanValue(array.opt(i)));
}
return res;
} else if (object instanceof String) {
String value = (String) object;
// XML 1.0 Allowed characters
// Char ::= #x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF]
return value
.codePoints()
.filter(
cp -> cp == 0x9 || cp == 0xA || cp == 0xD || (cp >= 0x20 && cp <= 0xD7FF)
|| (cp >= 0xE000 && cp <= 0xFFFD)
|| (cp >= 0x10000 && cp <= 0x10FFFF))
.collect(
StringBuilder::new,
StringBuilder::appendCodePoint,
StringBuilder::append)
.toString();
}
return object;
}
private static String cleanKey(String key) {
if (key == null || key.isEmpty()) {
return key;
}
// xml tag cannot begin with "-", ".", or a numeric digit.
switch (key.charAt(0)) {
case '-':
case '.':
key = "_" + key.substring(1);
break;
}
if (Character.isDigit(key.charAt(0))) {
if (key.matches("^[0-9]+$")) {
// add prefix of only numbers in JSON Keynames with 'm_'
key = "m_" + key;
} else {
// add prefix of startNumbers in JSON Keynames with 'n_'
key = "n_" + key;
}
}
StringBuilder res = new StringBuilder(key.length());
for (int i = 0; i < key.length(); i++) {
char c = key.charAt(i);
// sequence of whitespaces are rendered as a single '_'
if (Character.isWhitespace(c)) {
while (i + 1 < key.length() && Character.isWhitespace(key.charAt(i + 1))) {
i++;
}
res.append('_');
}
// remove invalid chars for xml tags with the expception of '=' and '/'
else if (INVALID_XMLTAG_CHARS.indexOf(c) >= 0) {
switch (c) {
case '=':
res.append('-');
break;
case '/':
res.append('_');
break;
default:
break;
}
// nothing
}
// all other chars are kept
else {
res.append(c);
}
}
return res.toString();
}
static public String convertToXML(final String jsonRecord) {
if (log.isTraceEnabled()) {
log.trace("input json: " + jsonRecord);
}
JSONObject jsonObject = cleanJsonObject(new org.json.JSONObject(jsonRecord));
String res = XML_HEADER + org.json.XML.toString(jsonObject, XML_WRAP_TAG); // wrap xml in single root element
if (log.isTraceEnabled()) {
log.trace("outout xml: " + res);
}
return res;
}
}

View File

@ -48,7 +48,7 @@ public class XSLTTransformationFunction implements MapFunction<MetadataRecord, M
@Override
public MetadataRecord call(MetadataRecord value) {
aggregationCounter.getTotalItems().add(1);
try {
Processor processor = new Processor(false);
processor.registerExtensionFunction(cleanFunction);
@ -60,11 +60,18 @@ public class XSLTTransformationFunction implements MapFunction<MetadataRecord, M
comp.setParameter(datasourceIDParam, new XdmAtomicValue(value.getProvenance().getDatasourceId()));
QName datasourceNameParam = new QName(DATASOURCE_NAME_PARAM);
comp.setParameter(datasourceNameParam, new XdmAtomicValue(value.getProvenance().getDatasourceName()));
XsltExecutable xslt = comp
.compile(new StreamSource(IOUtils.toInputStream(transformationRule, StandardCharsets.UTF_8)));
XdmNode source = processor
.newDocumentBuilder()
.build(new StreamSource(IOUtils.toInputStream(value.getBody(), StandardCharsets.UTF_8)));
XsltExecutable xslt;
XdmNode source;
try {
xslt = comp
.compile(new StreamSource(IOUtils.toInputStream(transformationRule, StandardCharsets.UTF_8)));
source = processor
.newDocumentBuilder()
.build(new StreamSource(IOUtils.toInputStream(value.getBody(), StandardCharsets.UTF_8)));
} catch (Throwable e) {
throw new RuntimeException("Error on parsing xslt", e);
}
try {
XsltTransformer trans = xslt.load();
trans.setInitialContextNode(source);
final StringWriter output = new StringWriter();

View File

@ -17,6 +17,12 @@
"paramDescription": "the path to get the input data from Pubmed",
"paramRequired": true
},
{
"paramName": "oip",
"paramLongName": "openapcInputPath",
"paramDescription": "the path to get the input data from OpenAPC",
"paramRequired": true
},
{
"paramName": "o",
"paramLongName": "outputPath",

View File

@ -31,6 +31,7 @@ spark2SqlQueryExecutionListeners=com.cloudera.spark.lineage.NavigatorQueryListen
# The following is needed as a property of a workflow
oozie.wf.application.path=${oozieTopWfApplicationPath}
crossrefInputPath=/data/bip-affiliations/data.json
crossrefInputPath=/data/bip-affiliations/crossref-data.json
pubmedInputPath=/data/bip-affiliations/pubmed-data.json
openapcInputPath=/data/bip-affiliations/openapc-data.json
outputPath=/tmp/crossref-affiliations-output-v5

View File

@ -9,6 +9,10 @@
<name>pubmedInputPath</name>
<description>the path where to find the inferred affiliation relations from Pubmed</description>
</property>
<property>
<name>openapcInputPath</name>
<description>the path where to find the inferred affiliation relations from OpenAPC</description>
</property>
<property>
<name>outputPath</name>
<description>the path where to store the actionset</description>
@ -102,6 +106,7 @@
</spark-opts>
<arg>--crossrefInputPath</arg><arg>${crossrefInputPath}</arg>
<arg>--pubmedInputPath</arg><arg>${pubmedInputPath}</arg>
<arg>--openapcInputPath</arg><arg>${openapcInputPath}</arg>
<arg>--outputPath</arg><arg>${outputPath}</arg>
</spark>
<ok to="End"/>

View File

@ -16,5 +16,10 @@
"paramLongName": "outputPath",
"paramDescription": "the path of the new ActionSet",
"paramRequired": true
}
}, {
"paramName": "fd",
"paramLongName": "distributeDoi",
"paramDescription": "the path of the new ActionSet",
"paramRequired": false
}
]

View File

@ -0,0 +1,20 @@
[
{
"paramName": "sp",
"paramLongName": "sourcePath",
"paramDescription": "the zipped opencitations file",
"paramRequired": true
},
{
"paramName": "op",
"paramLongName": "outputPath",
"paramDescription": "the working path",
"paramRequired": true
},
{
"paramName": "issm",
"paramLongName": "isSparkSessionManaged",
"paramDescription": "the hdfs name node",
"paramRequired": false
}
]

View File

@ -0,0 +1,30 @@
<configuration>
<property>
<name>jobTracker</name>
<value>yarnRM</value>
</property>
<property>
<name>nameNode</name>
<value>hdfs://nameservice1</value>
</property>
<property>
<name>oozie.use.system.libpath</name>
<value>true</value>
</property>
<property>
<name>hiveMetastoreUris</name>
<value>thrift://iis-cdh5-test-m3.ocean.icm.edu.pl:9083</value>
</property>
<property>
<name>hiveJdbcUrl</name>
<value>jdbc:hive2://iis-cdh5-test-m3.ocean.icm.edu.pl:10000</value>
</property>
<property>
<name>hiveDbName</name>
<value>openaire</value>
</property>
<property>
<name>oozie.launcher.mapreduce.user.classpath.first</name>
<value>true</value>
</property>
</configuration>

View File

@ -0,0 +1,153 @@
<workflow-app name="FOS no doi" xmlns="uri:oozie:workflow:0.5">
<parameters>
<property>
<name>fosPath</name>
<description>the input path of the resources to be extended</description>
</property>
<property>
<name>outputPath</name>
<description>the path where to store the actionset</description>
</property>
<property>
<name>sparkDriverMemory</name>
<description>memory for driver process</description>
</property>
<property>
<name>sparkExecutorMemory</name>
<description>memory for individual executor</description>
</property>
<property>
<name>sparkExecutorCores</name>
<description>number of cores used by single executor</description>
</property>
<property>
<name>oozieActionShareLibForSpark2</name>
<description>oozie action sharelib for spark 2.*</description>
</property>
<property>
<name>spark2ExtraListeners</name>
<value>com.cloudera.spark.lineage.NavigatorAppListener</value>
<description>spark 2.* extra listeners classname</description>
</property>
<property>
<name>spark2SqlQueryExecutionListeners</name>
<value>com.cloudera.spark.lineage.NavigatorQueryListener</value>
<description>spark 2.* sql query execution listeners classname</description>
</property>
<property>
<name>spark2YarnHistoryServerAddress</name>
<description>spark 2.* yarn history server address</description>
</property>
<property>
<name>spark2EventLogDir</name>
<description>spark 2.* event log dir location</description>
</property>
</parameters>
<global>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<configuration>
<property>
<name>mapreduce.job.queuename</name>
<value>${queueName}</value>
</property>
<property>
<name>oozie.launcher.mapred.job.queue.name</name>
<value>${oozieLauncherQueueName}</value>
</property>
<property>
<name>oozie.action.sharelib.for.spark</name>
<value>${oozieActionShareLibForSpark2}</value>
</property>
</configuration>
</global>
<start to="getFOS"/>
<kill name="Kill">
<message>Action failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
</kill>
<action name="getFOS">
<spark xmlns="uri:oozie:spark-action:0.2">
<master>yarn</master>
<mode>cluster</mode>
<name>Gets Data from FOS csv file</name>
<class>eu.dnetlib.dhp.actionmanager.createunresolvedentities.GetFOSSparkJob</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>--sourcePath</arg><arg>${fosPath}</arg>
<arg>--outputPath</arg><arg>${workingDir}/input/fos</arg>
<arg>--delimiter</arg><arg>${delimiter}</arg>
</spark>
<ok to="prepareFos"/>
<error to="Kill"/>
</action>
<action name="prepareFos">
<spark xmlns="uri:oozie:spark-action:0.2">
<master>yarn</master>
<mode>cluster</mode>
<name>Produces the results from FOS</name>
<class>eu.dnetlib.dhp.actionmanager.createunresolvedentities.PrepareFOSSparkJob</class>
<jar>dhp-aggregation-${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=${sparkSqlWarehouseDir}
</spark-opts>
<arg>--sourcePath</arg><arg>${workingDir}/input/fos</arg>
<arg>--outputPath</arg><arg>${workingDir}/prepared</arg>
<arg>--distributeDoi</arg><arg>false</arg>
</spark>
<ok to="produceActionSet"/>
<error to="Kill"/>
</action>
<action name="produceActionSet">
<spark xmlns="uri:oozie:spark-action:0.2">
<master>yarn</master>
<mode>cluster</mode>
<name>Save the action set grouping results with the same id</name>
<class>eu.dnetlib.dhp.actionmanager.fosnodoi.CreateActionSetSparkJob</class>
<jar>dhp-aggregation-${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=${sparkSqlWarehouseDir}
</spark-opts>
<arg>--sourcePath</arg><arg>${workingDir}/prepared/fos</arg>
<arg>--outputPath</arg><arg>${outputPath}</arg>
</spark>
<ok to="End"/>
<error to="Kill"/>
</action>
<end name="End"/>
</workflow-app>

View File

@ -1,13 +1,13 @@
[
{
"paramName": "if",
"paramLongName": "inputFile",
"paramName": "ip",
"paramLongName": "inputPath",
"paramDescription": "the zipped opencitations file",
"paramRequired": true
},
{
"paramName": "wp",
"paramLongName": "workingPath",
"paramName": "op",
"paramLongName": "outputPath",
"paramDescription": "the working path",
"paramRequired": true
},
@ -16,11 +16,5 @@
"paramLongName": "hdfsNameNode",
"paramDescription": "the hdfs name node",
"paramRequired": true
},
{
"paramName": "p",
"paramLongName": "prefix",
"paramDescription": "COCI or POCI",
"paramRequired": true
}
]

View File

@ -1,7 +1,7 @@
[
{
"paramName": "wp",
"paramLongName": "workingPath",
"paramName": "ip",
"paramLongName": "inputPath",
"paramDescription": "the zipped opencitations file",
"paramRequired": true
},
@ -24,15 +24,9 @@
"paramLongName": "outputPath",
"paramDescription": "the hdfs name node",
"paramRequired": true
},
{
"paramName": "if",
"paramLongName": "inputFile",
"paramDescription": "the hdfs name node",
"paramRequired": true
}, {
"paramName": "f",
"paramLongName": "format",
}, {
"paramName": "nn",
"paramLongName": "hdfsNameNode",
"paramDescription": "the hdfs name node",
"paramRequired": true
}

View File

@ -27,7 +27,9 @@
<case to="download">${wf:conf('resumeFrom') eq 'DownloadDump'}</case>
<case to="extract">${wf:conf('resumeFrom') eq 'ExtractContent'}</case>
<case to="read">${wf:conf('resumeFrom') eq 'ReadContent'}</case>
<default to="create_actionset"/> <!-- first action to be done when downloadDump is to be performed -->
<case to="remap">${wf:conf('resumeFrom') eq 'MapContent'}</case>
<case to="create_actionset">${wf:conf('resumeFrom') eq 'CreateAS'}</case>
<default to="deleteoutputpath"/> <!-- first action to be done when downloadDump is to be performed -->
</switch>
</decision>
@ -35,6 +37,15 @@
<message>Action failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
</kill>
<action name="deleteoutputpath">
<fs>
<delete path='${inputPath}'/>
<mkdir path='${inputPath}'/>
</fs>
<ok to="download"/>
<error to="Kill"/>
</action>
<action name="download">
<shell xmlns="uri:oozie:shell-action:0.2">
<job-tracker>${jobTracker}</job-tracker>
@ -47,7 +58,28 @@
</configuration>
<exec>download.sh</exec>
<argument>${filelist}</argument>
<argument>${workingPath}/${prefix}/Original</argument>
<argument>${inputPath}/Original</argument>
<env-var>HADOOP_USER_NAME=${wf:user()}</env-var>
<file>download.sh</file>
<capture-output/>
</shell>
<ok to="download_correspondence"/>
<error to="Kill"/>
</action>
<!-- downloads the correspondence from the omid and the pid (doi, pmid etc)-->
<action name="download_correspondence">
<shell xmlns="uri:oozie:shell-action:0.2">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<configuration>
<property>
<name>mapred.job.queue.name</name>
<value>${queueName}</value>
</property>
</configuration>
<exec>download_corr.sh</exec>
<argument>${filecorrespondence}</argument>
<argument>${inputPath}/correspondence</argument>
<env-var>HADOOP_USER_NAME=${wf:user()}</env-var>
<file>download.sh</file>
<capture-output/>
@ -60,9 +92,19 @@
<java>
<main-class>eu.dnetlib.dhp.actionmanager.opencitations.GetOpenCitationsRefs</main-class>
<arg>--hdfsNameNode</arg><arg>${nameNode}</arg>
<arg>--inputFile</arg><arg>${inputFile}</arg>
<arg>--workingPath</arg><arg>${workingPath}/${prefix}</arg>
<arg>--prefix</arg><arg>${prefix}</arg>
<arg>--inputPath</arg><arg>${inputPath}/Original</arg>
<arg>--outputPath</arg><arg>${inputPath}/Extracted</arg>
</java>
<ok to="read"/>
<error to="Kill"/>
</action>
<action name="extract_correspondence">
<java>
<main-class>eu.dnetlib.dhp.actionmanager.opencitations.GetOpenCitationsRefs</main-class>
<arg>--hdfsNameNode</arg><arg>${nameNode}</arg>
<arg>--inputPath</arg><arg>${inputPath}/correspondence</arg>
<arg>--outputPath</arg><arg>${inputPath}/correspondence_extracted</arg>
</java>
<ok to="read"/>
<error to="Kill"/>
@ -85,11 +127,35 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.sql.warehouse.dir=${sparkSqlWarehouseDir}
</spark-opts>
<arg>--workingPath</arg><arg>${workingPath}/${prefix}/${prefix}</arg>
<arg>--outputPath</arg><arg>${workingPath}/${prefix}/${prefix}_JSON/</arg>
<arg>--inputPath</arg><arg>${inputPath}/Extracted</arg>
<arg>--outputPath</arg><arg>${inputPath}/JSON</arg>
<arg>--delimiter</arg><arg>${delimiter}</arg>
<arg>--inputFile</arg><arg>${inputFileCoci}</arg>
<arg>--format</arg><arg>${prefix}</arg>
<arg>--hdfsNameNode</arg><arg>${nameNode}</arg>
</spark>
<ok to="remap"/>
<error to="Kill"/>
</action>
<action name="remap">
<spark xmlns="uri:oozie:spark-action:0.2">
<master>yarn</master>
<mode>cluster</mode>
<name>Produces the AS for OC</name>
<class>eu.dnetlib.dhp.actionmanager.opencitations.MapOCIdsInPids</class>
<jar>dhp-aggregation-${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=${sparkSqlWarehouseDir}
</spark-opts>
<arg>--inputPath</arg><arg>${inputPath}</arg>
<arg>--outputPath</arg><arg>${outputPathExtraction}</arg>
<arg>--nameNode</arg><arg>${nameNode}</arg>
</spark>
<ok to="create_actionset"/>
<error to="Kill"/>
@ -112,7 +178,7 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.sql.warehouse.dir=${sparkSqlWarehouseDir}
</spark-opts>
<arg>--inputPath</arg><arg>${workingPath}</arg>
<arg>--inputPath</arg><arg>${outputPathExtraction}</arg>
<arg>--outputPath</arg><arg>${outputPath}</arg>
</spark>
<ok to="End"/>

View File

@ -0,0 +1,25 @@
[
{
"paramName": "ip",
"paramLongName": "inputPath",
"paramDescription": "the zipped opencitations file",
"paramRequired": true
},
{
"paramName": "op",
"paramLongName": "outputPath",
"paramDescription": "the working path",
"paramRequired": true
},
{
"paramName": "issm",
"paramLongName": "isSparkSessionManged",
"paramDescription": "the hdfs name node",
"paramRequired": false
},{
"paramName": "nn",
"paramLongName": "nameNode",
"paramDescription": "the hdfs name node",
"paramRequired": true
}
]

View File

@ -0,0 +1,20 @@
[
{
"paramName": "ip",
"paramLongName": "inputPath",
"paramDescription": "the zipped opencitations file",
"paramRequired": true
},
{
"paramName": "op",
"paramLongName": "outputPath",
"paramDescription": "the working path",
"paramRequired": true
},
{
"paramName": "issm",
"paramLongName": "isSparkSessionManaged",
"paramDescription": "the hdfs name node",
"paramRequired": false
}
]

View File

@ -0,0 +1,30 @@
[
{
"paramName": "issm",
"paramLongName": "isSparkSessionManaged",
"paramDescription": "the hdfs name node",
"paramRequired": false
},
{
"paramName": "d",
"paramLongName": "delimiter",
"paramDescription": "the hdfs name node",
"paramRequired": false
},
{
"paramName": "op",
"paramLongName": "outputPath",
"paramDescription": "the hdfs name node",
"paramRequired": true
},
{
"paramName": "if",
"paramLongName": "inputFile",
"paramDescription": "the hdfs name node",
"paramRequired": true
}
]

View File

@ -0,0 +1,58 @@
<configuration>
<property>
<name>jobTracker</name>
<value>yarnRM</value>
</property>
<property>
<name>nameNode</name>
<value>hdfs://nameservice1</value>
</property>
<property>
<name>oozie.use.system.libpath</name>
<value>true</value>
</property>
<property>
<name>oozie.action.sharelib.for.spark</name>
<value>spark2</value>
</property>
<property>
<name>hive_metastore_uris</name>
<value>thrift://iis-cdh5-test-m3.ocean.icm.edu.pl:9083</value>
</property>
<property>
<name>spark2YarnHistoryServerAddress</name>
<value>http://iis-cdh5-test-gw.ocean.icm.edu.pl:18089</value>
</property>
<property>
<name>spark2ExtraListeners</name>
<value>com.cloudera.spark.lineage.NavigatorAppListener</value>
</property>
<property>
<name>spark2SqlQueryExecutionListeners</name>
<value>com.cloudera.spark.lineage.NavigatorQueryListener</value>
</property>
<property>
<name>oozie.launcher.mapreduce.user.classpath.first</name>
<value>true</value>
</property>
<property>
<name>sparkExecutorNumber</name>
<value>4</value>
</property>
<property>
<name>spark2EventLogDir</name>
<value>/user/spark/spark2ApplicationHistory</value>
</property>
<property>
<name>sparkDriverMemory</name>
<value>15G</value>
</property>
<property>
<name>sparkExecutorMemory</name>
<value>6G</value>
</property>
<property>
<name>sparkExecutorCores</name>
<value>1</value>
</property>
</configuration>

View File

@ -0,0 +1,2 @@
#!/bin/bash
curl -L $1 | hdfs dfs -put - $2

View File

@ -0,0 +1,82 @@
<workflow-app name="Transfomative Agreement Integration" xmlns="uri:oozie:workflow:0.5">
<global>
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<configuration>
<property>
<name>mapreduce.job.queuename</name>
<value>${queueName}</value>
</property>
<property>
<name>oozie.launcher.mapred.job.queue.name</name>
<value>${oozieLauncherQueueName}</value>
</property>
<property>
<name>oozie.action.sharelib.for.spark</name>
<value>${oozieActionShareLibForSpark2}</value>
</property>
</configuration>
</global>
<start to="resume_from"/>
<decision name="resume_from">
<switch>
<case to="download">${wf:conf('resumeFrom') eq 'DownloadDump'}</case>
<default to="create_actionset"/> <!-- first action to be done when downloadDump is to be performed -->
</switch>
</decision>
<kill name="Kill">
<message>Action failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
</kill>
<action name="download">
<shell xmlns="uri:oozie:shell-action:0.2">
<job-tracker>${jobTracker}</job-tracker>
<name-node>${nameNode}</name-node>
<configuration>
<property>
<name>mapred.job.queue.name</name>
<value>${queueName}</value>
</property>
</configuration>
<exec>download.sh</exec>
<argument>${inputFile}</argument>
<argument>${workingDir}/transformativeagreement/transformativeAgreement.json</argument>
<env-var>HADOOP_USER_NAME=${wf:user()}</env-var>
<file>download.sh</file>
<capture-output/>
</shell>
<ok to="create_actionset"/>
<error to="Kill"/>
</action>
<action name="create_actionset">
<spark xmlns="uri:oozie:spark-action:0.2">
<master>yarn</master>
<mode>cluster</mode>
<name>Produces the AS for the Transformative Agreement</name>
<class>eu.dnetlib.dhp.actionmanager.transformativeagreement.CreateActionSetSparkJob</class>
<jar>dhp-aggregation-${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=${sparkSqlWarehouseDir}
</spark-opts>
<arg>--inputPath</arg><arg>${workingDir}/transformativeagreement/</arg>
<arg>--outputPath</arg><arg>${outputPath}</arg>
</spark>
<ok to="End"/>
<error to="Kill"/>
</action>
<end name="End"/>
</workflow-app>

View File

@ -0,0 +1,114 @@
BEGIN;
INSERT INTO dsm_services(
_dnet_resource_identifier_,
id,
officialname,
englishname,
namespaceprefix,
websiteurl,
logourl,
platform,
contactemail,
collectedfrom,
provenanceaction,
_typology_to_remove_,
eosc_type,
eosc_datasource_type,
research_entity_types,
thematic
) VALUES (
'openaire____::base_search',
'openaire____::base_search',
'Bielefeld Academic Search Engine (BASE)',
'Bielefeld Academic Search Engine (BASE)',
'base_search_',
'https://www.base-search.net',
'https://www.base-search.net/about/download/logo_224x57_white.gif',
'BASE',
'openaire-helpdesk@uni-bielefeld.de',
'infrastruct_::openaire',
'user:insert',
'aggregator::pubsrepository::unknown',
'Data Source',
'Aggregator',
ARRAY['Research Products'],
false
);
INSERT INTO dsm_service_organization(
_dnet_resource_identifier_,
organization,
service
) VALUES (
'fairsharing_::org::214@@openaire____::base_search',
'fairsharing_::org::214',
'openaire____::base_search'
);
INSERT INTO dsm_api(
_dnet_resource_identifier_,
id,
service,
protocol,
baseurl,
metadata_identifier_path
) VALUES (
'api_________::openaire____::base_search::dump',
'api_________::openaire____::base_search::dump',
'openaire____::base_search',
'baseDump',
'/user/michele.artini/base-import/base_oaipmh_dump-current.tar',
'//*[local-name()=''header'']/*[local-name()=''identifier'']'
);
INSERT INTO dsm_apiparams(
_dnet_resource_identifier_,
api,
param,
value
) VALUES (
'api_________::openaire____::base_search::dump@@dbUrl',
'api_________::openaire____::base_search::dump',
'dbUrl',
'jdbc:postgresql://postgresql.services.openaire.eu:5432/dnet_openaireplus'
);
INSERT INTO dsm_apiparams(
_dnet_resource_identifier_,
api,
param,
value
) VALUES (
'api_________::openaire____::base_search::dump@@dbUser',
'api_________::openaire____::base_search::dump',
'dbUser',
'dnet'
);
INSERT INTO dsm_apiparams(
_dnet_resource_identifier_,
api,
param,
value
) VALUES (
'api_________::openaire____::base_search::dump@@dbPassword',
'api_________::openaire____::base_search::dump',
'dbPassword',
'***'
);
INSERT INTO dsm_apiparams(
_dnet_resource_identifier_,
api,
param,
value
) VALUES (
'api_________::openaire____::base_search::dump@@acceptedNormTypes',
'api_________::openaire____::base_search::dump',
'acceptedNormTypes',
'1,11,111,121,14,15,18,181,182,183,1A,6,7'
);
COMMIT;

View File

@ -0,0 +1,9 @@
select s.id as id
from dsm_services s
where collectedfrom = 'openaire____::opendoar'
and jurisdiction = 'Institutional'
and s.id in (
select service from dsm_api where coalesce(compatibility_override, compatibility) = 'driver' or coalesce(compatibility_override, compatibility) = 'UNKNOWN'
) and s.id not in (
select service from dsm_api where coalesce(compatibility_override, compatibility) like '%openaire%'
);

View File

@ -0,0 +1,11 @@
select
s.id as id,
s.jurisdiction as jurisdiction,
array_remove(array_agg(a.id || ' (compliance: ' || coalesce(a.compatibility_override, a.compatibility, 'UNKNOWN') || ')@@@' || coalesce(a.last_collection_total, 0)), NULL) as aggregations
from
dsm_services s
join dsm_api a on (s.id = a.service)
where
collectedfrom = 'openaire____::opendoar'
group by
s.id;

View File

@ -0,0 +1,180 @@
<RESOURCE_PROFILE>
<HEADER>
<RESOURCE_IDENTIFIER value="c67911d6-9988-4a3b-b965-7d39bdd4a31d_Vm9jYWJ1bGFyeURTUmVzb3VyY2VzL1ZvY2FidWxhcnlEU1Jlc291cmNlVHlwZQ==" />
<RESOURCE_TYPE value="VocabularyDSResourceType" />
<RESOURCE_KIND value="VocabularyDSResources" />
<RESOURCE_URI value="" />
<DATE_OF_CREATION value="2024-02-13T11:15:48+00:00" />
</HEADER>
<BODY>
<CONFIGURATION>
<VOCABULARY_NAME code="base:normalized_types">base:normalized_types</VOCABULARY_NAME>
<VOCABULARY_DESCRIPTION>base:normalized_types</VOCABULARY_DESCRIPTION>
<TERMS>
<TERM native_name="Text" code="Text" english_name="Text" encoding="BASE">
<SYNONYMS>
<SYNONYM term="1" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Book" code="Book" english_name="Book" encoding="BASE">
<SYNONYMS>
<SYNONYM term="11" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Book part" code="Book part" english_name="Book part" encoding="BASE">
<SYNONYMS>
<SYNONYM term="111" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Journal/Newspaper" code="Journal/Newspaper" english_name="Journal/Newspaper" encoding="BASE">
<SYNONYMS>
<SYNONYM term="12" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Article contribution" code="Article contribution" english_name="Article contribution" encoding="BASE">
<SYNONYMS>
<SYNONYM term="121" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Other non-article" code="Other non-article" english_name="Other non-article" encoding="BASE">
<SYNONYMS>
<SYNONYM term="122" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Conference object" code="Conference object" english_name="Conference object" encoding="BASE">
<SYNONYMS>
<SYNONYM term="13" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Report" code="Report" english_name="Report" encoding="BASE">
<SYNONYMS>
<SYNONYM term="14" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Review" code="Review" english_name="Review" encoding="BASE">
<SYNONYMS>
<SYNONYM term="15" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Course material" code="Course material" english_name="Course material" encoding="BASE">
<SYNONYMS>
<SYNONYM term="16" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Lecture" code="Lecture" english_name="Lecture" encoding="BASE">
<SYNONYMS>
<SYNONYM term="17" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Thesis" code="Thesis" english_name="Thesis" encoding="BASE">
<SYNONYMS>
<SYNONYM term="18" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Bachelor's thesis" code="Bachelor's thesis" english_name="Bachelor's thesis" encoding="BASE">
<SYNONYMS>
<SYNONYM term="181" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Master's thesis" code="Master's thesis" english_name="Master's thesis" encoding="BASE">
<SYNONYMS>
<SYNONYM term="182" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Doctoral and postdoctoral thesis" code="Doctoral and postdoctoral thesis" english_name="Doctoral and postdoctoral thesis" encoding="BASE">
<SYNONYMS>
<SYNONYM term="183" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Manuscript" code="Manuscript" english_name="Manuscript" encoding="BASE">
<SYNONYMS>
<SYNONYM term="19" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Patent" code="Patent" english_name="Patent" encoding="BASE">
<SYNONYMS>
<SYNONYM term="1A" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Musical notation" code="Musical notation" english_name="Musical notation" encoding="BASE">
<SYNONYMS>
<SYNONYM term="2" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Map" code="Map" english_name="Map" encoding="BASE">
<SYNONYMS>
<SYNONYM term="3" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Audio" code="Audio" english_name="Audio" encoding="BASE">
<SYNONYMS>
<SYNONYM term="4" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Image/Video" code="Image/Video" english_name="Image/Video" encoding="BASE">
<SYNONYMS>
<SYNONYM term="5" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Still image" code="Still image" english_name="Still image" encoding="BASE">
<SYNONYMS>
<SYNONYM term="51" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Moving image/Video" code="Moving image/Video" english_name="Moving image/Video" encoding="BASE">
<SYNONYMS>
<SYNONYM term="52" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Software" code="Software" english_name="Software" encoding="BASE">
<SYNONYMS>
<SYNONYM term="6" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Dataset" code="Dataset" english_name="Dataset" encoding="BASE">
<SYNONYMS>
<SYNONYM term="7" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
<TERM native_name="Unknown" code="Unknown" english_name="Unknown" encoding="BASE">
<SYNONYMS>
<SYNONYM term="F" encoding="BASE" />
</SYNONYMS>
<RELATIONS />
</TERM>
</TERMS>
</CONFIGURATION>
<STATUS>
<LAST_UPDATE value="2013-11-18T10:46:36Z" />
</STATUS>
<SECURITY_PARAMETERS>String</SECURITY_PARAMETERS>
</BODY>
</RESOURCE_PROFILE>

View File

@ -0,0 +1,432 @@
<RESOURCE_PROFILE>
<HEADER>
<RESOURCE_IDENTIFIER value="" />
<RESOURCE_TYPE value="TransformationRuleDSResourceType" />
<RESOURCE_KIND value="TransformationRuleDSResources" />
<RESOURCE_URI value="" />
<DATE_OF_CREATION value="2024-03-05T11:23:00+00:00" />
</HEADER>
<BODY>
<CONFIGURATION>
<SOURCE_METADATA_FORMAT interpretation="cleaned" layout="store" name="dc" />
<SINK_METADATA_FORMAT name="oaf_hbase" />
<IMPORTED />
<SCRIPT>
<TITLE>xslt_base2oaf_hadoop</TITLE>
<CODE>
<xsl:stylesheet xmlns:oaire="http://namespace.openaire.eu/schema/oaire/" xmlns:dateCleaner="http://eu/dnetlib/transform/dateISO"
xmlns:base_dc="http://oai.base-search.net/base_dc/"
xmlns:datacite="http://datacite.org/schema/kernel-4" xmlns:dr="http://www.driver-repository.eu/namespace/dr" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform" xmlns:vocabulary="http://eu/dnetlib/transform/clean" xmlns:oaf="http://namespace.openaire.eu/oaf"
xmlns:oai="http://www.openarchives.org/OAI/2.0/" xmlns:dri="http://www.driver-repository.eu/namespace/dri" xmlns:xs="http://www.w3.org/2001/XMLSchema" xmlns:dc="http://purl.org/dc/elements/1.1/"
exclude-result-prefixes="xsl vocabulary dateCleaner base_dc" version="2.0">
<xsl:param name="varOfficialName" />
<xsl:param name="varDataSourceId" />
<xsl:param name="varFP7" select="'corda_______::'" />
<xsl:param name="varH2020" select="'corda__h2020::'" />
<xsl:param name="repoCode" select="substring-before(//*[local-name() = 'header']/*[local-name()='recordIdentifier'], ':')" />
<xsl:param name="index" select="0" />
<xsl:param name="transDate" select="current-dateTime()" />
<xsl:template name="terminate">
<xsl:message terminate="yes">
record is not compliant, transformation is interrupted.
</xsl:message>
</xsl:template>
<xsl:template match="/">
<record>
<xsl:apply-templates select="//*[local-name() = 'header']" />
<!-- TO EVALUATE
base_dc:authod_id
base_dc:authod_id/base_dc:creator_id
base_dc:authod_id/base_dc:creator_name
example:
<dc:creator>ALBU, Svetlana</dc:creator>
<base_dc:authod_id>
<base_dc:creator_name>ALBU, Svetlana</base_dc:creator_name>
<base_dc:creator_id>https://orcid.org/0000-0002-8648-950X</base_dc:creator_id>
</base_dc:authod_id>
-->
<!-- NOT USED
base_dc:global_id (I used oai:identifier)
base_dc:collection/text()
base_dc:continent
base_dc:country
base_dc:year (I used dc:date)
dc:coverage
dc:language (I used base_dc:lang)
base_dc:link (I used dc:identifier)
-->
<metadata>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:title" />
<xsl:with-param name="targetElement" select="'dc:title'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:creator/replace(., '^(.*)\|.*$', '$1')" />
<xsl:with-param name="targetElement" select="'dc:creator'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:contributor" />
<xsl:with-param name="targetElement" select="'dc:contributor'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:description" />
<xsl:with-param name="targetElement" select="'dc:description'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:subject" />
<xsl:with-param name="targetElement" select="'dc:subject'" />
</xsl:call-template>
<!-- TODO: I'm not sure if this is the correct encoding -->
<xsl:for-each select="//base_dc:classcode|//base_dc:autoclasscode">
<dc:subject><xsl:value-of select="concat(@type, ':', .)" /></dc:subject>
</xsl:for-each>
<!-- END TODO -->
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:publisher" />
<xsl:with-param name="targetElement" select="'dc:publisher'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:format" />
<xsl:with-param name="targetElement" select="'dc:format'" />
</xsl:call-template>
<xsl:for-each select="//base_dc:typenorm">
<dc:type>
<xsl:value-of select="vocabulary:clean(., 'base:normalized_types')" />
</dc:type>
</xsl:for-each>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:type" />
<xsl:with-param name="targetElement" select="'dc:type'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:source" />
<xsl:with-param name="targetElement" select="'dc:source'" />
</xsl:call-template>
<dc:language>
<xsl:value-of select="vocabulary:clean( //base_dc:lang, 'dnet:languages')" />
</dc:language>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:rights" />
<xsl:with-param name="targetElement" select="'dc:rights'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:relation" />
<xsl:with-param name="targetElement" select="'dc:relation'" />
</xsl:call-template>
<xsl:if test="not(//dc:identifier[starts-with(., 'http')])">
<xsl:call-template name="terminate" />
</xsl:if>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:identifier[starts-with(., 'http')]" />
<xsl:with-param name="targetElement" select="'dc:identifier'" />
</xsl:call-template>
<xsl:for-each select="//dc:relation">
<xsl:if test="matches(normalize-space(.), '(info:eu-repo/grantagreement/ec/fp7/)(\d\d\d\d\d\d)(.*)', 'i')">
<oaf:projectid>
<xsl:value-of select="concat($varFP7, replace(normalize-space(.), '(info:eu-repo/grantagreement/ec/fp7/)(\d\d\d\d\d\d)(.*)', '$2', 'i'))" />
</oaf:projectid>
</xsl:if>
<xsl:if test="matches(normalize-space(.), '(info:eu-repo/grantagreement/ec/h2020/)(\d\d\d\d\d\d)(.*)', 'i')">
<oaf:projectid>
<xsl:value-of select="concat($varH2020, replace(normalize-space(.), '(info:eu-repo/grantagreement/ec/h2020/)(\d\d\d\d\d\d)(.*)', '$2', 'i'))" />
</oaf:projectid>
</xsl:if>
</xsl:for-each>
<xsl:choose>
<!-- I used an inline mapping because the field typenorm could be repeated and I have to specify a list of priority -->
<!-- Book part -->
<xsl:when test="//base_dc:typenorm = '111'">
<dr:CobjCategory type="publication">0013</dr:CobjCategory>
</xsl:when>
<!-- Book -->
<xsl:when test="//base_dc:typenorm = '11'">
<dr:CobjCategory type="publication">0002</dr:CobjCategory>
</xsl:when>
<!-- Article contribution -->
<xsl:when test="//base_dc:typenorm = '121'">
<dr:CobjCategory type="publication">0001</dr:CobjCategory>
</xsl:when>
<!-- Journal/Newspaper -->
<xsl:when test="//base_dc:typenorm = '12'">
<dr:CobjCategory type="publication">0043</dr:CobjCategory>
</xsl:when>
<!-- Report -->
<xsl:when test="//base_dc:typenorm = '14'">
<dr:CobjCategory type="publication">0017</dr:CobjCategory>
</xsl:when>
<!-- Review -->
<xsl:when test="//base_dc:typenorm = '15'">
<dr:CobjCategory type="publication">0015</dr:CobjCategory>
</xsl:when>
<!-- Lecture -->
<xsl:when test="//base_dc:typenorm = '17'">
<dr:CobjCategory type="publication">0010</dr:CobjCategory>
</xsl:when>
<!-- Bachelor's thesis -->
<xsl:when test="//base_dc:typenorm = '181'">
<dr:CobjCategory type="publication">0008</dr:CobjCategory>
</xsl:when>
<!-- Master's thesis -->
<xsl:when test="//base_dc:typenorm = '182'">
<dr:CobjCategory type="publication">0007</dr:CobjCategory>
</xsl:when>
<!-- Doctoral and postdoctoral thesis -->
<xsl:when test="//base_dc:typenorm = '183'">
<dr:CobjCategory type="publication">0006</dr:CobjCategory>
</xsl:when>
<!-- Thesis -->
<xsl:when test="//base_dc:typenorm = '18'">
<dr:CobjCategory type="publication">0044</dr:CobjCategory>
</xsl:when>
<!-- Patent -->
<xsl:when test="//base_dc:typenorm = '1A'">
<dr:CobjCategory type="publication">0019</dr:CobjCategory>
</xsl:when>
<!-- Text -->
<xsl:when test="//base_dc:typenorm = '1'">
<dr:CobjCategory type="publication">0001</dr:CobjCategory>
</xsl:when>
<!-- Software -->
<xsl:when test="//base_dc:typenorm = '6'">
<dr:CobjCategory type="software">0029</dr:CobjCategory>
</xsl:when>
<!-- Dataset -->
<xsl:when test="//base_dc:typenorm = '7'">
<dr:CobjCategory type="dataset">0021</dr:CobjCategory>
</xsl:when>
<!-- Still image -->
<xsl:when test="//base_dc:typenorm = '51'">
<dr:CobjCategory type="other">0025</dr:CobjCategory>
</xsl:when>
<!-- Moving image/Video -->
<xsl:when test="//base_dc:typenorm = '52'">
<dr:CobjCategory type="other">0024</dr:CobjCategory>
</xsl:when>
<!-- Image/Video -->
<xsl:when test="//base_dc:typenorm = '5'">
<dr:CobjCategory type="other">0033</dr:CobjCategory>
</xsl:when>
<!-- Audio -->
<xsl:when test="//base_dc:typenorm = '4'">
<dr:CobjCategory type="other">0030</dr:CobjCategory>
</xsl:when>
<!-- Musical notation -->
<xsl:when test="//base_dc:typenorm = '2'">
<dr:CobjCategory type="other">0020</dr:CobjCategory>
</xsl:when>
<!-- Map -->
<xsl:when test="//base_dc:typenorm = '3'">
<dr:CobjCategory type="other">0020</dr:CobjCategory>
</xsl:when>
<!-- Other non-article -->
<xsl:when test="//base_dc:typenorm = '122'">
<dr:CobjCategory type="publication">0038</dr:CobjCategory>
</xsl:when>
<!-- Course material -->
<xsl:when test="//base_dc:typenorm = '16'">
<dr:CobjCategory type="publication">0038</dr:CobjCategory>
</xsl:when>
<!-- Manuscript -->
<xsl:when test="//base_dc:typenorm = '19'">
<dr:CobjCategory type="publication">0038</dr:CobjCategory>
</xsl:when>
<!-- Conference object -->
<xsl:when test="//base_dc:typenorm = '13'">
<dr:CobjCategory type="publication">0004</dr:CobjCategory>
</xsl:when>
<!-- Unknown -->
<xsl:when test="//base_dc:typenorm = 'F'">
<dr:CobjCategory type="other">0000</dr:CobjCategory>
</xsl:when>
<xsl:otherwise>
<dr:CobjCategory type="other">0000</dr:CobjCategory>
</xsl:otherwise>
</xsl:choose>
<oaf:accessrights>
<xsl:choose>
<xsl:when test="//base_dc:oa[.='0']">CLOSED</xsl:when>
<xsl:when test="//base_dc:oa[.='1']">OPEN</xsl:when>
<xsl:when test="//base_dc:oa[.='2']">UNKNOWN</xsl:when>
<xsl:when test="//base_dc:rightsnorm">
<xsl:value-of select="vocabulary:clean(//base_dc:rightsnorm, 'dnet:access_modes')" />
</xsl:when>
<xsl:when test="//dc:rights">
<xsl:value-of select="vocabulary:clean( //dc:rights, 'dnet:access_modes')" />
</xsl:when>
<xsl:otherwise>UNKNOWN</xsl:otherwise>
</xsl:choose>
</oaf:accessrights>
<xsl:for-each select="//base_dc:doi">
<oaf:identifier identifierType="doi">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:for-each select="distinct-values(//dc:identifier[starts-with(., 'http') and (not(contains(., '://dx.doi.org/') or contains(., '://doi.org/') or contains(., '://hdl.handle.net/')))])">
<oaf:identifier identifierType="url">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:for-each select="distinct-values(//dc:identifier[starts-with(., 'http') and contains(., '://hdl.handle.net/')]/substring-after(., 'hdl.handle.net/'))">
<oaf:identifier identifierType="handle">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:for-each select="distinct-values(//dc:identifier[starts-with(., 'urn:nbn:nl:') or starts-with(., 'URN:NBN:NL:')])">
<oaf:identifier identifierType='urn'>
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<oaf:identifier identifierType="oai-original">
<xsl:value-of
select="//oai:header/oai:identifier" />
</oaf:identifier>
<oaf:hostedBy>
<xsl:attribute name="name">
<xsl:value-of select="//base_dc:collname" />
</xsl:attribute>
<xsl:attribute name="id">
<xsl:value-of select="concat('opendoar____::', //base_dc:collection/@opendoar_id)" />
</xsl:attribute>
</oaf:hostedBy>
<oaf:collectedFrom>
<xsl:attribute name="name">
<xsl:value-of select="$varOfficialName" />
</xsl:attribute>
<xsl:attribute name="id">
<xsl:value-of select="$varDataSourceId" />
</xsl:attribute>
</oaf:collectedFrom>
<oaf:dateAccepted>
<xsl:value-of select="dateCleaner:dateISO( //dc:date[1] )" />
</oaf:dateAccepted>
<xsl:if test="//base_dc:oa[.='1']">
<xsl:for-each select="//dc:relation[starts-with(., 'http')]">
<oaf:fulltext>
<xsl:value-of select="normalize-space(.)" />
</oaf:fulltext>
</xsl:for-each>
</xsl:if>
<xsl:for-each select="//base_dc:collection/@ror_id">
<oaf:relation relType="resultOrganization"
subRelType="affiliation"
relClass="hasAuthorInstitution"
targetType="organization">
<xsl:choose>
<xsl:when test="contains(.,'https://ror.org/')">
<xsl:value-of select="concat('ror_________::', normalize-space(.))" />
</xsl:when>
<xsl:otherwise>
<xsl:value-of select="concat('ror_________::https://ror.org/', normalize-space(.))" />
</xsl:otherwise>
</xsl:choose>
</oaf:relation>
</xsl:for-each>
</metadata>
<xsl:copy-of select="//*[local-name() = 'about']" />
</record>
</xsl:template>
<xsl:template name="allElements">
<xsl:param name="sourceElement" />
<xsl:param name="targetElement" />
<xsl:for-each select="$sourceElement">
<xsl:element name="{$targetElement}">
<xsl:value-of select="normalize-space(.)" />
</xsl:element>
</xsl:for-each>
</xsl:template>
<xsl:template match="//*[local-name() = 'header']">
<xsl:if test="//oai:header/@status='deleted'">
<xsl:call-template name="terminate" />
</xsl:if>
<xsl:copy>
<xsl:apply-templates select="node()|@*" />
<xsl:element name="dr:dateOfTransformation">
<xsl:value-of select="$transDate" />
</xsl:element>
</xsl:copy>
</xsl:template>
<xsl:template match="node()|@*">
<xsl:copy>
<xsl:apply-templates select="node()|@*" />
</xsl:copy>
</xsl:template>
</xsl:stylesheet>
</CODE>
</SCRIPT>
</CONFIGURATION>
<STATUS />
<SECURITY_PARAMETERS />
</BODY>
</RESOURCE_PROFILE>

View File

@ -0,0 +1,461 @@
<RESOURCE_PROFILE>
<HEADER>
<RESOURCE_IDENTIFIER value="2ad0cdd9-c96c-484c-8b0e-ed56d86891fe_VHJhbnNmb3JtYXRpb25SdWxlRFNSZXNvdXJjZXMvVHJhbnNmb3JtYXRpb25SdWxlRFNSZXNvdXJjZVR5cGU=" />
<RESOURCE_TYPE value="TransformationRuleDSResourceType" />
<RESOURCE_KIND value="TransformationRuleDSResources" />
<RESOURCE_URI value="" />
<DATE_OF_CREATION value="2024-03-05T11:23:00+00:00" />
</HEADER>
<BODY>
<CONFIGURATION>
<SOURCE_METADATA_FORMAT interpretation="cleaned" layout="store" name="dc" />
<SINK_METADATA_FORMAT name="odf_hbase" />
<IMPORTED />
<SCRIPT>
<TITLE>xslt_base2odf_hadoop</TITLE>
<CODE>
<xsl:stylesheet xmlns:oaire="http://namespace.openaire.eu/schema/oaire/" xmlns:dateCleaner="http://eu/dnetlib/transform/dateISO" xmlns:base_dc="http://oai.base-search.net/base_dc/"
xmlns:datacite="http://datacite.org/schema/kernel-4" xmlns:dr="http://www.driver-repository.eu/namespace/dr" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform" xmlns:vocabulary="http://eu/dnetlib/transform/clean" xmlns:oaf="http://namespace.openaire.eu/oaf"
xmlns:oai="http://www.openarchives.org/OAI/2.0/" xmlns:dri="http://www.driver-repository.eu/namespace/dri" xmlns:xs="http://www.w3.org/2001/XMLSchema" xmlns:dc="http://purl.org/dc/elements/1.1/"
exclude-result-prefixes="xsl vocabulary dateCleaner base_dc" version="2.0">
<xsl:param name="varOfficialName" />
<xsl:param name="varDataSourceId" />
<xsl:param name="varFP7" select="'corda_______::'" />
<xsl:param name="varH2020" select="'corda__h2020::'" />
<xsl:param name="repoCode" select="substring-before(//*[local-name() = 'header']/*[local-name()='recordIdentifier'], ':')" />
<xsl:param name="index" select="0" />
<xsl:param name="transDate" select="current-dateTime()" />
<xsl:template name="terminate">
<xsl:message terminate="yes">
record is not compliant, transformation is interrupted.
</xsl:message>
</xsl:template>
<xsl:template match="/">
<record>
<xsl:apply-templates select="//*[local-name() = 'header']" />
<!-- NOT USED
base_dc:global_id (I used oai:identifier)
base_dc:collection/text()
base_dc:continent
base_dc:country
dc:coverage
dc:source
dc:relation
dc:type (I used //base_dc:typenorm)
dc:language (I used base_dc:lang)
base_dc:link (I used dc:identifier)
-->
<metadata>
<datacite:resource>
<xsl:for-each select="//base_dc:doi">
<datacite:identifier identifierType="DOI">
<xsl:value-of select="." />
</datacite:identifier>
</xsl:for-each>
<datacite:alternateIdentifiers>
<xsl:for-each
select="distinct-values(//dc:identifier[starts-with(., 'http') and (not(contains(., '://dx.doi.org/') or contains(., '://doi.org/') or contains(., '://hdl.handle.net/')))])">
<datacite:identifier alternateIdentifierType="url">
<xsl:value-of select="." />
</datacite:identifier>
</xsl:for-each>
<xsl:for-each select="distinct-values(//dc:identifier[starts-with(., 'http') and contains(., '://hdl.handle.net/')]/substring-after(., 'hdl.handle.net/'))">
<datacite:identifier alternateIdentifierType="handle">
<xsl:value-of select="." />
</datacite:identifier>
</xsl:for-each>
<xsl:for-each select="distinct-values(//dc:identifier[starts-with(., 'urn:nbn:nl:') or starts-with(., 'URN:NBN:NL:')])">
<datacite:identifier alternateIdentifierType='urn'>
<xsl:value-of select="." />
</datacite:identifier>
</xsl:for-each>
<datacite:identifier alternateIdentifierType="oai-original">
<xsl:value-of
select="//oai:header/oai:identifier" />
</datacite:identifier>
</datacite:alternateIdentifiers>
<datacite:relatedIdentifiers />
<xsl:for-each select="//base_dc:typenorm">
<datacite:resourceType><xsl:value-of select="vocabulary:clean(., 'base:normalized_types')" /></datacite:resourceType>
</xsl:for-each>
<datacite:titles>
<xsl:for-each select="//dc:title">
<datacite:title>
<xsl:value-of select="normalize-space(.)" />
</datacite:title>
</xsl:for-each>
</datacite:titles>
<datacite:creators>
<xsl:for-each select="//dc:creator">
<xsl:variable name="author" select="normalize-space(.)" />
<datacite:creator>
<datacite:creatorName>
<xsl:value-of select="$author" />
</datacite:creatorName>
<xsl:for-each select="//base_dc:authod_id[normalize-space(./base_dc:creator_name) = $author]/base_dc:creator_id ">
<xsl:if test="contains(.,'https://orcid.org/')">
<nameIdentifier schemeURI="https://orcid.org/" nameIdentifierScheme="ORCID">
<xsl:value-of select="substring-after(., 'https://orcid.org/')" />
</nameIdentifier>
</xsl:if>
</xsl:for-each>
</datacite:creator>
</xsl:for-each>
</datacite:creators>
<datacite:contributors>
<xsl:for-each select="//dc:contributor">
<datacite:contributor>
<datacite:contributorName>
<xsl:value-of select="normalize-space(.)" />
</datacite:contributorName>
</datacite:contributor>
</xsl:for-each>
</datacite:contributors>
<datacite:descriptions>
<xsl:for-each select="//dc:description">
<datacite:description descriptionType="Abstract">
<xsl:value-of select="normalize-space(.)" />
</datacite:description>
</xsl:for-each>
</datacite:descriptions>
<datacite:subjects>
<xsl:for-each select="//dc:subject">
<datacite:subject>
<xsl:value-of select="normalize-space(.)" />
</datacite:subject>
</xsl:for-each>
<xsl:for-each select="//base_dc:classcode|//base_dc:autoclasscode">
<datacite:subject subjectScheme="{@type}" classificationCode="{normalize-space(.)}">
<!-- TODO the value should be obtained by the Code -->
<xsl:value-of select="normalize-space(.)" />
</datacite:subject>
</xsl:for-each>
</datacite:subjects>
<xsl:for-each select="//dc:publisher">
<datacite:publisher>
<xsl:value-of select="normalize-space(.)" />
</datacite:publisher>
</xsl:for-each>
<xsl:for-each select="//base_dc:year">
<datacite:publicationYear>
<xsl:value-of select="normalize-space(.)" />
</datacite:publicationYear>
</xsl:for-each>
<datacite:formats>
<xsl:for-each select="//dc:format">
<datacite:format>
<xsl:value-of select="normalize-space(.)" />
</datacite:format>
</xsl:for-each>
</datacite:formats>
<datacite:language>
<xsl:value-of select="vocabulary:clean( //base_dc:lang, 'dnet:languages')" />
</datacite:language>
<oaf:accessrights>
<xsl:if test="//base_dc:oa[.='0']">
<datacite:rights rightsURI="http://purl.org/coar/access_right/c_16ec">restricted access</datacite:rights>
</xsl:if>
<xsl:if test="//base_dc:oa[.='1']">
<datacite:rights rightsURI="http://purl.org/coar/access_right/c_abf2">open access</datacite:rights>
</xsl:if>
<xsl:for-each select="//dc:rights|//base_dc:rightsnorm">
<datacite:rights><xsl:value-of select="vocabulary:clean(., 'dnet:access_modes')" /></datacite:rights>
</xsl:for-each>
</oaf:accessrights>
</datacite:resource>
<xsl:for-each select="//dc:relation">
<xsl:if test="matches(normalize-space(.), '(info:eu-repo/grantagreement/ec/fp7/)(\d\d\d\d\d\d)(.*)', 'i')">
<oaf:projectid>
<xsl:value-of select="concat($varFP7, replace(normalize-space(.), '(info:eu-repo/grantagreement/ec/fp7/)(\d\d\d\d\d\d)(.*)', '$2', 'i'))" />
</oaf:projectid>
</xsl:if>
<xsl:if test="matches(normalize-space(.), '(info:eu-repo/grantagreement/ec/h2020/)(\d\d\d\d\d\d)(.*)', 'i')">
<oaf:projectid>
<xsl:value-of select="concat($varH2020, replace(normalize-space(.), '(info:eu-repo/grantagreement/ec/h2020/)(\d\d\d\d\d\d)(.*)', '$2', 'i'))" />
</oaf:projectid>
</xsl:if>
</xsl:for-each>
<xsl:choose>
<!-- I used an inline mapping because the field typenorm could be repeated and I have to specify a list of priority -->
<!-- Book part -->
<xsl:when test="//base_dc:typenorm = '111'">
<dr:CobjCategory type="publication">0013</dr:CobjCategory>
</xsl:when>
<!-- Book -->
<xsl:when test="//base_dc:typenorm = '11'">
<dr:CobjCategory type="publication">0002</dr:CobjCategory>
</xsl:when>
<!-- Article contribution -->
<xsl:when test="//base_dc:typenorm = '121'">
<dr:CobjCategory type="publication">0001</dr:CobjCategory>
</xsl:when>
<!-- Journal/Newspaper -->
<xsl:when test="//base_dc:typenorm = '12'">
<dr:CobjCategory type="publication">0043</dr:CobjCategory>
</xsl:when>
<!-- Report -->
<xsl:when test="//base_dc:typenorm = '14'">
<dr:CobjCategory type="publication">0017</dr:CobjCategory>
</xsl:when>
<!-- Review -->
<xsl:when test="//base_dc:typenorm = '15'">
<dr:CobjCategory type="publication">0015</dr:CobjCategory>
</xsl:when>
<!-- Lecture -->
<xsl:when test="//base_dc:typenorm = '17'">
<dr:CobjCategory type="publication">0010</dr:CobjCategory>
</xsl:when>
<!-- Bachelor's thesis -->
<xsl:when test="//base_dc:typenorm = '181'">
<dr:CobjCategory type="publication">0008</dr:CobjCategory>
</xsl:when>
<!-- Master's thesis -->
<xsl:when test="//base_dc:typenorm = '182'">
<dr:CobjCategory type="publication">0007</dr:CobjCategory>
</xsl:when>
<!-- Doctoral and postdoctoral thesis -->
<xsl:when test="//base_dc:typenorm = '183'">
<dr:CobjCategory type="publication">0006</dr:CobjCategory>
</xsl:when>
<!-- Thesis -->
<xsl:when test="//base_dc:typenorm = '18'">
<dr:CobjCategory type="publication">0044</dr:CobjCategory>
</xsl:when>
<!-- Patent -->
<xsl:when test="//base_dc:typenorm = '1A'">
<dr:CobjCategory type="publication">0019</dr:CobjCategory>
</xsl:when>
<!-- Text -->
<xsl:when test="//base_dc:typenorm = '1'">
<dr:CobjCategory type="publication">0001</dr:CobjCategory>
</xsl:when>
<!-- Software -->
<xsl:when test="//base_dc:typenorm = '6'">
<dr:CobjCategory type="software">0029</dr:CobjCategory>
</xsl:when>
<!-- Dataset -->
<xsl:when test="//base_dc:typenorm = '7'">
<dr:CobjCategory type="dataset">0021</dr:CobjCategory>
</xsl:when>
<!-- Still image -->
<xsl:when test="//base_dc:typenorm = '51'">
<dr:CobjCategory type="other">0025</dr:CobjCategory>
</xsl:when>
<!-- Moving image/Video -->
<xsl:when test="//base_dc:typenorm = '52'">
<dr:CobjCategory type="other">0024</dr:CobjCategory>
</xsl:when>
<!-- Image/Video -->
<xsl:when test="//base_dc:typenorm = '5'">
<dr:CobjCategory type="other">0033</dr:CobjCategory>
</xsl:when>
<!-- Audio -->
<xsl:when test="//base_dc:typenorm = '4'">
<dr:CobjCategory type="other">0030</dr:CobjCategory>
</xsl:when>
<!-- Musical notation -->
<xsl:when test="//base_dc:typenorm = '2'">
<dr:CobjCategory type="other">0020</dr:CobjCategory>
</xsl:when>
<!-- Map -->
<xsl:when test="//base_dc:typenorm = '3'">
<dr:CobjCategory type="other">0020</dr:CobjCategory>
</xsl:when>
<!-- Other non-article -->
<xsl:when test="//base_dc:typenorm = '122'">
<dr:CobjCategory type="publication">0038</dr:CobjCategory>
</xsl:when>
<!-- Course material -->
<xsl:when test="//base_dc:typenorm = '16'">
<dr:CobjCategory type="publication">0038</dr:CobjCategory>
</xsl:when>
<!-- Manuscript -->
<xsl:when test="//base_dc:typenorm = '19'">
<dr:CobjCategory type="publication">0038</dr:CobjCategory>
</xsl:when>
<!-- Conference object -->
<xsl:when test="//base_dc:typenorm = '13'">
<dr:CobjCategory type="publication">0004</dr:CobjCategory>
</xsl:when>
<!-- Unknown -->
<xsl:when test="//base_dc:typenorm = 'F'">
<dr:CobjCategory type="other">0000</dr:CobjCategory>
</xsl:when>
<xsl:otherwise>
<dr:CobjCategory type="other">0000</dr:CobjCategory>
</xsl:otherwise>
</xsl:choose>
<oaf:accessrights>
<xsl:choose>
<xsl:when test="//base_dc:oa[.='0']">CLOSED</xsl:when>
<xsl:when test="//base_dc:oa[.='1']">OPEN</xsl:when>
<xsl:when test="//base_dc:oa[.='2']">UNKNOWN</xsl:when>
<xsl:when test="//base_dc:rightsnorm">
<xsl:value-of select="vocabulary:clean(//base_dc:rightsnorm, 'dnet:access_modes')" />
</xsl:when>
<xsl:when test="//dc:rights">
<xsl:value-of select="vocabulary:clean( //dc:rights, 'dnet:access_modes')" />
</xsl:when>
<xsl:otherwise>UNKNOWN</xsl:otherwise>
</xsl:choose>
</oaf:accessrights>
<xsl:for-each select="//base_dc:doi">
<oaf:identifier identifierType="doi">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:for-each
select="distinct-values(//dc:identifier[starts-with(., 'http') and ( not(contains(., '://dx.doi.org/') or contains(., '://doi.org/') or contains(., '://hdl.handle.net/')))])">
<oaf:identifier identifierType="url">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:for-each select="distinct-values(//dc:identifier[starts-with(., 'http') and contains(., '://hdl.handle.net/')]/substring-after(., 'hdl.handle.net/'))">
<oaf:identifier identifierType="handle">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:for-each select="distinct-values(//dc:identifier[starts-with(., 'urn:nbn:nl:') or starts-with(., 'URN:NBN:NL:')])">
<oaf:identifier identifierType='urn'>
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<oaf:identifier identifierType="oai-original">
<xsl:value-of
select="//oai:header/oai:identifier" />
</oaf:identifier>
<oaf:hostedBy>
<xsl:attribute name="name">
<xsl:value-of select="//base_dc:collname" />
</xsl:attribute>
<xsl:attribute name="id">
<xsl:value-of select="concat('opendoar____::', //base_dc:collection/@opendoar_id)" />
</xsl:attribute>
</oaf:hostedBy>
<oaf:collectedFrom>
<xsl:attribute name="name">
<xsl:value-of select="$varOfficialName" />
</xsl:attribute>
<xsl:attribute name="id">
<xsl:value-of select="$varDataSourceId" />
</xsl:attribute>
</oaf:collectedFrom>
<oaf:dateAccepted>
<xsl:value-of select="dateCleaner:dateISO( //dc:date[1] )" />
</oaf:dateAccepted>
<xsl:if test="//base_dc:oa[.='1']">
<xsl:for-each select="//dc:relation[starts-with(., 'http')]">
<oaf:fulltext>
<xsl:value-of select="normalize-space(.)" />
</oaf:fulltext>
</xsl:for-each>
</xsl:if>
<xsl:for-each select="//base_dc:collection/@ror_id">
<oaf:relation relType="resultOrganization" subRelType="affiliation" relClass="hasAuthorInstitution" targetType="organization">
<xsl:choose>
<xsl:when test="contains(.,'https://ror.org/')">
<xsl:value-of select="concat('ror_________::', normalize-space(.))" />
</xsl:when>
<xsl:otherwise>
<xsl:value-of select="concat('ror_________::https://ror.org/', normalize-space(.))" />
</xsl:otherwise>
</xsl:choose>
</oaf:relation>
</xsl:for-each>
</metadata>
<xsl:copy-of select="//*[local-name() = 'about']" />
</record>
</xsl:template>
<xsl:template match="//*[local-name() = 'header']">
<xsl:if test="//oai:header/@status='deleted'">
<xsl:call-template name="terminate" />
</xsl:if>
<xsl:copy>
<xsl:apply-templates select="node()|@*" />
<xsl:element name="dr:dateOfTransformation">
<xsl:value-of select="$transDate" />
</xsl:element>
</xsl:copy>
</xsl:template>
<xsl:template match="node()|@*">
<xsl:copy>
<xsl:apply-templates select="node()|@*" />
</xsl:copy>
</xsl:template>
</xsl:stylesheet>
</CODE>
</SCRIPT>
</CONFIGURATION>
<STATUS />
<SECURITY_PARAMETERS />
</BODY>
</RESOURCE_PROFILE>

View File

@ -78,10 +78,6 @@ public class PrepareAffiliationRelationsTest {
.getResource("/eu/dnetlib/dhp/actionmanager/bipaffiliations/doi_to_ror.json")
.getPath();
String pubmedAffiliationRelationsPath = getClass()
.getResource("/eu/dnetlib/dhp/actionmanager/bipaffiliations/doi_to_ror.json")
.getPath();
String outputPath = workingDir.toString() + "/actionSet";
PrepareAffiliationRelations
@ -89,7 +85,8 @@ public class PrepareAffiliationRelationsTest {
new String[] {
"-isSparkSessionManaged", Boolean.FALSE.toString(),
"-crossrefInputPath", crossrefAffiliationRelationPath,
"-pubmedInputPath", pubmedAffiliationRelationsPath,
"-pubmedInputPath", crossrefAffiliationRelationPath,
"-openapcInputPath", crossrefAffiliationRelationPath,
"-outputPath", outputPath
});
@ -106,7 +103,7 @@ public class PrepareAffiliationRelationsTest {
// );
// }
// count the number of relations
assertEquals(40, tmp.count());
assertEquals(60, tmp.count());
Dataset<Relation> dataset = spark.createDataset(tmp.rdd(), Encoders.bean(Relation.class));
dataset.createOrReplaceTempView("result");
@ -117,7 +114,7 @@ public class PrepareAffiliationRelationsTest {
// verify that we have equal number of bi-directional relations
Assertions
.assertEquals(
20, execVerification
30, execVerification
.filter(
"relClass='" + ModelConstants.HAS_AUTHOR_INSTITUTION + "'")
.collectAsList()
@ -125,7 +122,7 @@ public class PrepareAffiliationRelationsTest {
Assertions
.assertEquals(
20, execVerification
30, execVerification
.filter(
"relClass='" + ModelConstants.IS_AUTHOR_INSTITUTION_OF + "'")
.collectAsList()

View File

@ -0,0 +1,104 @@
package eu.dnetlib.dhp.actionmanager.fosnodoi;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import org.apache.commons.io.FileUtils;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.LocalFileSystem;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.SparkSession;
import org.junit.jupiter.api.*;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.createunresolvedentities.GetFOSSparkJob;
import eu.dnetlib.dhp.actionmanager.createunresolvedentities.PrepareTest;
import eu.dnetlib.dhp.actionmanager.createunresolvedentities.ProduceTest;
import eu.dnetlib.dhp.actionmanager.createunresolvedentities.model.FOSDataModel;
/**
* @author miriam.baglioni
* @Date 13/02/23
*/
public class GetFosTest {
private static final Logger log = LoggerFactory.getLogger(ProduceTest.class);
private static Path workingDir;
private static SparkSession spark;
private static LocalFileSystem fs;
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
@BeforeAll
public static void beforeAll() throws IOException {
workingDir = Files.createTempDirectory(PrepareTest.class.getSimpleName());
fs = FileSystem.getLocal(new Configuration());
log.info("using work dir {}", workingDir);
SparkConf conf = new SparkConf();
conf.setAppName(ProduceTest.class.getSimpleName());
conf.setMaster("local[*]");
conf.set("spark.driver.host", "localhost");
conf.set("hive.metastore.local", "true");
conf.set("spark.ui.enabled", "false");
conf.set("spark.sql.warehouse.dir", workingDir.toString());
conf.set("hive.metastore.warehouse.dir", workingDir.resolve("warehouse").toString());
spark = SparkSession
.builder()
.appName(PrepareTest.class.getSimpleName())
.config(conf)
.getOrCreate();
}
@AfterAll
public static void afterAll() throws IOException {
FileUtils.deleteDirectory(workingDir.toFile());
spark.stop();
}
@Test
@Disabled
void test3() throws Exception {
final String sourcePath = getClass()
.getResource("/eu/dnetlib/dhp/actionmanager/fosnodoi/fosnodoi.csv")
.getPath();
final String outputPath = workingDir.toString() + "/fos.json";
GetFOSSparkJob
.main(
new String[] {
"--isSparkSessionManaged", Boolean.FALSE.toString(),
"--sourcePath", sourcePath,
"-outputPath", outputPath,
"-delimiter", ","
});
final JavaSparkContext sc = JavaSparkContext.fromSparkContext(spark.sparkContext());
JavaRDD<FOSDataModel> tmp = sc
.textFile(outputPath)
.map(item -> OBJECT_MAPPER.readValue(item, FOSDataModel.class));
tmp.foreach(t -> Assertions.assertTrue(t.getOaid() != null));
tmp.foreach(t -> Assertions.assertTrue(t.getLevel1() != null));
tmp.foreach(t -> Assertions.assertTrue(t.getLevel2() != null));
tmp.foreach(t -> Assertions.assertTrue(t.getLevel3() != null));
tmp.foreach(t -> System.out.println(new ObjectMapper().writeValueAsString(t)));
}
}

View File

@ -0,0 +1,99 @@
package eu.dnetlib.dhp.actionmanager.fosnodoi;
import static org.junit.jupiter.api.Assertions.assertEquals;
import static org.junit.jupiter.api.Assertions.assertTrue;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import org.apache.commons.io.FileUtils;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.LocalFileSystem;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.SparkSession;
import org.junit.jupiter.api.AfterAll;
import org.junit.jupiter.api.Assertions;
import org.junit.jupiter.api.BeforeAll;
import org.junit.jupiter.api.Test;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.createunresolvedentities.PrepareFOSSparkJob;
import eu.dnetlib.dhp.actionmanager.createunresolvedentities.PrepareSDGSparkJob;
import eu.dnetlib.dhp.actionmanager.createunresolvedentities.ProduceTest;
import eu.dnetlib.dhp.schema.oaf.Result;
public class PrepareTest {
private static final Logger log = LoggerFactory.getLogger(ProduceTest.class);
private static Path workingDir;
private static SparkSession spark;
private static LocalFileSystem fs;
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
@BeforeAll
public static void beforeAll() throws IOException {
workingDir = Files.createTempDirectory(PrepareTest.class.getSimpleName());
fs = FileSystem.getLocal(new Configuration());
log.info("using work dir {}", workingDir);
SparkConf conf = new SparkConf();
conf.setAppName(ProduceTest.class.getSimpleName());
conf.setMaster("local[*]");
conf.set("spark.driver.host", "localhost");
conf.set("hive.metastore.local", "true");
conf.set("spark.ui.enabled", "false");
conf.set("spark.sql.warehouse.dir", workingDir.toString());
conf.set("hive.metastore.warehouse.dir", workingDir.resolve("warehouse").toString());
spark = SparkSession
.builder()
.appName(PrepareTest.class.getSimpleName())
.config(conf)
.getOrCreate();
}
@AfterAll
public static void afterAll() throws IOException {
FileUtils.deleteDirectory(workingDir.toFile());
spark.stop();
}
@Test
void fosPrepareTest() throws Exception {
final String sourcePath = getClass()
.getResource("/eu/dnetlib/dhp/actionmanager/fosnodoi/fosnodoi.json")
.getPath();
PrepareFOSSparkJob
.main(
new String[] {
"--isSparkSessionManaged", Boolean.FALSE.toString(),
"--sourcePath", sourcePath,
"-outputPath", workingDir.toString() + "/work",
"-distributeDoi", Boolean.FALSE.toString()
});
final JavaSparkContext sc = JavaSparkContext.fromSparkContext(spark.sparkContext());
JavaRDD<Result> tmp = sc
.textFile(workingDir.toString() + "/work/fos")
.map(item -> OBJECT_MAPPER.readValue(item, Result.class));
tmp.foreach(t -> System.out.println(new ObjectMapper().writeValueAsString(t)));
}
}

View File

@ -76,7 +76,7 @@ public class CreateOpenCitationsASTest {
String inputPath = getClass()
.getResource(
"/eu/dnetlib/dhp/actionmanager/opencitations/COCI")
"/eu/dnetlib/dhp/actionmanager/opencitations/COCI/inputremap/jsonforas")
.getPath();
CreateActionSetSparkJob
@ -84,8 +84,6 @@ public class CreateOpenCitationsASTest {
new String[] {
"-isSparkSessionManaged",
Boolean.FALSE.toString(),
"-shouldDuplicateRels",
Boolean.TRUE.toString(),
"-inputPath",
inputPath,
"-outputPath",
@ -99,9 +97,10 @@ public class CreateOpenCitationsASTest {
.map(value -> OBJECT_MAPPER.readValue(value._2().toString(), AtomicAction.class))
.map(aa -> ((Relation) aa.getPayload()));
assertEquals(31, tmp.count());
Assertions.assertEquals(27, tmp.count());
tmp.foreach(r -> Assertions.assertEquals(1, r.getCollectedfrom().size()));
// tmp.foreach(r -> System.out.println(OBJECT_MAPPER.writeValueAsString(r)));
tmp.foreach(r -> System.out.println(OBJECT_MAPPER.writeValueAsString(r)));
}

View File

@ -0,0 +1,90 @@
package eu.dnetlib.dhp.actionmanager.opencitations;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import org.apache.commons.io.FileUtils;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.LocalFileSystem;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.Encoders;
import org.apache.spark.sql.SparkSession;
import org.junit.jupiter.api.AfterAll;
import org.junit.jupiter.api.Assertions;
import org.junit.jupiter.api.BeforeAll;
import org.junit.jupiter.api.Test;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.opencitations.model.COCI;
/**
* @author miriam.baglioni
* @Date 07/03/24
*/
public class RemapTest {
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
private static SparkSession spark;
private static Path workingDir;
private static final Logger log = LoggerFactory
.getLogger(RemapTest.class);
@BeforeAll
public static void beforeAll() throws IOException {
workingDir = Files
.createTempDirectory(RemapTest.class.getSimpleName());
log.info("using work dir {}", workingDir);
SparkConf conf = new SparkConf();
conf.setAppName(RemapTest.class.getSimpleName());
conf.setMaster("local[*]");
conf.set("spark.driver.host", "localhost");
conf.set("hive.metastore.local", "true");
conf.set("spark.ui.enabled", "false");
conf.set("spark.sql.warehouse.dir", workingDir.toString());
conf.set("hive.metastore.warehouse.dir", workingDir.resolve("warehouse").toString());
spark = SparkSession
.builder()
.appName(RemapTest.class.getSimpleName())
.config(conf)
.getOrCreate();
}
@AfterAll
public static void afterAll() throws IOException {
FileUtils.deleteDirectory(workingDir.toFile());
spark.stop();
}
@Test
void testRemap() throws Exception {
String inputPath = getClass()
.getResource(
"/eu/dnetlib/dhp/actionmanager/opencitations/COCI/inputremap")
.getPath();
MapOCIdsInPids
.main(
new String[] {
"-isSparkSessionManged",
Boolean.FALSE.toString(),
"-inputPath",
inputPath,
"-outputPath",
workingDir.toString() + "/out/",
"-nameNode", "input1;input2;input3;input4;input5"
});
}
}

View File

@ -0,0 +1,324 @@
package eu.dnetlib.dhp.actionmanager.transformativeagreement;
import static org.junit.jupiter.api.Assertions.assertEquals;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import org.apache.commons.io.FileUtils;
import org.apache.hadoop.io.Text;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.SparkSession;
import org.junit.jupiter.api.AfterAll;
import org.junit.jupiter.api.BeforeAll;
import org.junit.jupiter.api.Test;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.actionmanager.opencitations.CreateActionSetSparkJob;
import eu.dnetlib.dhp.actionmanager.opencitations.CreateOpenCitationsASTest;
import eu.dnetlib.dhp.schema.action.AtomicAction;
import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.oaf.Relation;
import eu.dnetlib.dhp.schema.oaf.utils.CleaningFunctions;
import eu.dnetlib.dhp.schema.oaf.utils.IdentifierFactory;
/**
* @author miriam.baglioni
* @Date 13/02/24
*/
public class CreateTAActionSetTest {
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
private static SparkSession spark;
private static Path workingDir;
private static final Logger log = LoggerFactory
.getLogger(CreateOpenCitationsASTest.class);
@BeforeAll
public static void beforeAll() throws IOException {
workingDir = Files
.createTempDirectory(CreateTAActionSetTest.class.getSimpleName());
log.info("using work dir {}", workingDir);
SparkConf conf = new SparkConf();
conf.setAppName(CreateTAActionSetTest.class.getSimpleName());
conf.setMaster("local[*]");
conf.set("spark.driver.host", "localhost");
conf.set("hive.metastore.local", "true");
conf.set("spark.ui.enabled", "false");
conf.set("spark.sql.warehouse.dir", workingDir.toString());
conf.set("hive.metastore.warehouse.dir", workingDir.resolve("warehouse").toString());
spark = SparkSession
.builder()
.appName(CreateTAActionSetTest.class.getSimpleName())
.config(conf)
.getOrCreate();
}
@AfterAll
public static void afterAll() throws IOException {
FileUtils.deleteDirectory(workingDir.toFile());
spark.stop();
}
@Test
void createActionSet() throws Exception {
String inputPath = getClass()
.getResource(
"/eu/dnetlib/dhp/actionmanager/transformativeagreement/facts.json")
.getPath();
eu.dnetlib.dhp.actionmanager.transformativeagreement.CreateActionSetSparkJob
.main(
new String[] {
"-isSparkSessionManaged",
Boolean.FALSE.toString(),
"-inputPath",
inputPath,
"-outputPath",
workingDir.toString() + "/actionSet1"
});
}
@Test
void testNumberofRelations2() throws Exception {
String inputPath = getClass()
.getResource(
"/eu/dnetlib/dhp/actionmanager/opencitations/COCI")
.getPath();
eu.dnetlib.dhp.actionmanager.opencitations.CreateActionSetSparkJob
.main(
new String[] {
"-isSparkSessionManaged",
Boolean.FALSE.toString(),
"-inputPath",
inputPath,
"-outputPath",
workingDir.toString() + "/actionSet2"
});
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
JavaRDD<Relation> tmp = sc
.sequenceFile(workingDir.toString() + "/actionSet2", Text.class, Text.class)
.map(value -> OBJECT_MAPPER.readValue(value._2().toString(), AtomicAction.class))
.map(aa -> ((Relation) aa.getPayload()));
assertEquals(23, tmp.count());
// tmp.foreach(r -> System.out.println(OBJECT_MAPPER.writeValueAsString(r)));
}
@Test
void testRelationsCollectedFrom() throws Exception {
String inputPath = getClass()
.getResource(
"/eu/dnetlib/dhp/actionmanager/opencitations/COCI")
.getPath();
eu.dnetlib.dhp.actionmanager.opencitations.CreateActionSetSparkJob
.main(
new String[] {
"-isSparkSessionManaged",
Boolean.FALSE.toString(),
"-inputPath",
inputPath,
"-outputPath",
workingDir.toString() + "/actionSet3"
});
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
JavaRDD<Relation> tmp = sc
.sequenceFile(workingDir.toString() + "/actionSet3", Text.class, Text.class)
.map(value -> OBJECT_MAPPER.readValue(value._2().toString(), AtomicAction.class))
.map(aa -> ((Relation) aa.getPayload()));
tmp.foreach(r -> {
assertEquals(ModelConstants.OPENOCITATIONS_NAME, r.getCollectedfrom().get(0).getValue());
assertEquals(ModelConstants.OPENOCITATIONS_ID, r.getCollectedfrom().get(0).getKey());
});
}
@Test
void testRelationsDataInfo() throws Exception {
String inputPath = getClass()
.getResource(
"/eu/dnetlib/dhp/actionmanager/opencitations/COCI")
.getPath();
eu.dnetlib.dhp.actionmanager.opencitations.CreateActionSetSparkJob
.main(
new String[] {
"-isSparkSessionManaged",
Boolean.FALSE.toString(),
"-inputPath",
inputPath,
"-outputPath",
workingDir.toString() + "/actionSet4"
});
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
JavaRDD<Relation> tmp = sc
.sequenceFile(workingDir.toString() + "/actionSet4", Text.class, Text.class)
.map(value -> OBJECT_MAPPER.readValue(value._2().toString(), AtomicAction.class))
.map(aa -> ((Relation) aa.getPayload()));
tmp.foreach(r -> {
assertEquals(false, r.getDataInfo().getInferred());
assertEquals(false, r.getDataInfo().getDeletedbyinference());
assertEquals("0.91", r.getDataInfo().getTrust());
assertEquals(
eu.dnetlib.dhp.actionmanager.opencitations.CreateActionSetSparkJob.OPENCITATIONS_CLASSID,
r.getDataInfo().getProvenanceaction().getClassid());
assertEquals(
eu.dnetlib.dhp.actionmanager.opencitations.CreateActionSetSparkJob.OPENCITATIONS_CLASSNAME,
r.getDataInfo().getProvenanceaction().getClassname());
assertEquals(ModelConstants.DNET_PROVENANCE_ACTIONS, r.getDataInfo().getProvenanceaction().getSchemeid());
assertEquals(ModelConstants.DNET_PROVENANCE_ACTIONS, r.getDataInfo().getProvenanceaction().getSchemename());
});
}
@Test
void testRelationsSemantics() throws Exception {
String inputPath = getClass()
.getResource(
"/eu/dnetlib/dhp/actionmanager/opencitations/COCI")
.getPath();
eu.dnetlib.dhp.actionmanager.opencitations.CreateActionSetSparkJob
.main(
new String[] {
"-isSparkSessionManaged",
Boolean.FALSE.toString(),
"-inputPath",
inputPath,
"-outputPath",
workingDir.toString() + "/actionSet5"
});
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
JavaRDD<Relation> tmp = sc
.sequenceFile(workingDir.toString() + "/actionSet5", Text.class, Text.class)
.map(value -> OBJECT_MAPPER.readValue(value._2().toString(), AtomicAction.class))
.map(aa -> ((Relation) aa.getPayload()));
tmp.foreach(r -> {
assertEquals("citation", r.getSubRelType());
assertEquals("resultResult", r.getRelType());
});
assertEquals(23, tmp.filter(r -> r.getRelClass().equals("Cites")).count());
assertEquals(0, tmp.filter(r -> r.getRelClass().equals("IsCitedBy")).count());
}
@Test
void testRelationsSourceTargetPrefix() throws Exception {
String inputPath = getClass()
.getResource(
"/eu/dnetlib/dhp/actionmanager/opencitations/COCI")
.getPath();
eu.dnetlib.dhp.actionmanager.opencitations.CreateActionSetSparkJob
.main(
new String[] {
"-isSparkSessionManaged",
Boolean.FALSE.toString(),
"-inputPath",
inputPath,
"-outputPath",
workingDir.toString() + "/actionSet6"
});
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
JavaRDD<Relation> tmp = sc
.sequenceFile(workingDir.toString() + "/actionSet6", Text.class, Text.class)
.map(value -> OBJECT_MAPPER.readValue(value._2().toString(), AtomicAction.class))
.map(aa -> ((Relation) aa.getPayload()));
tmp.foreach(r -> {
assertEquals("50|doi_________::", r.getSource().substring(0, 17));
assertEquals("50|doi_________::", r.getTarget().substring(0, 17));
});
}
@Test
void testRelationsSourceTargetCouple() throws Exception {
final String doi1 = "50|doi_________::"
+ IdentifierFactory.md5(CleaningFunctions.normalizePidValue("doi", "10.1007/s10854-015-3684-x"));
final String doi2 = "50|doi_________::"
+ IdentifierFactory.md5(CleaningFunctions.normalizePidValue("doi", "10.1111/j.1551-2916.2008.02408.x"));
final String doi3 = "50|doi_________::"
+ IdentifierFactory.md5(CleaningFunctions.normalizePidValue("doi", "10.1007/s10854-014-2114-9"));
final String doi4 = "50|doi_________::"
+ IdentifierFactory.md5(CleaningFunctions.normalizePidValue("doi", "10.1016/j.ceramint.2013.09.069"));
final String doi5 = "50|doi_________::"
+ IdentifierFactory.md5(CleaningFunctions.normalizePidValue("doi", "10.1007/s10854-009-9913-4"));
final String doi6 = "50|doi_________::"
+ IdentifierFactory.md5(CleaningFunctions.normalizePidValue("doi", "10.1016/0038-1098(72)90370-5"));
String inputPath = getClass()
.getResource(
"/eu/dnetlib/dhp/actionmanager/opencitations/COCI")
.getPath();
CreateActionSetSparkJob
.main(
new String[] {
"-isSparkSessionManaged",
Boolean.FALSE.toString(),
"-inputPath",
inputPath,
"-outputPath",
workingDir.toString() + "/actionSet7"
});
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
JavaRDD<Relation> tmp = sc
.sequenceFile(workingDir.toString() + "/actionSet7", Text.class, Text.class)
.map(value -> OBJECT_MAPPER.readValue(value._2().toString(), AtomicAction.class))
.map(aa -> ((Relation) aa.getPayload()));
JavaRDD<Relation> check = tmp.filter(r -> r.getSource().equals(doi1) || r.getTarget().equals(doi1));
assertEquals(5, check.count());
// check.foreach(r -> {
// if (r.getSource().equals(doi2) || r.getSource().equals(doi3) || r.getSource().equals(doi4) ||
// r.getSource().equals(doi5) || r.getSource().equals(doi6)) {
// assertEquals(ModelConstants.IS_CITED_BY, r.getRelClass());
// assertEquals(doi1, r.getTarget());
// }
// });
assertEquals(5, check.filter(r -> r.getSource().equals(doi1)).count());
check.filter(r -> r.getSource().equals(doi1)).foreach(r -> assertEquals(ModelConstants.CITES, r.getRelClass()));
}
}

View File

@ -0,0 +1,38 @@
package eu.dnetlib.dhp.collection.plugin.base;
import java.io.Serializable;
public class BaseCollectionInfo implements Serializable {
private static final long serialVersionUID = 5766333937429419647L;
private String id;
private String opendoarId;
private String rorId;
public String getId() {
return this.id;
}
public void setId(final String id) {
this.id = id;
}
public String getOpendoarId() {
return this.opendoarId;
}
public void setOpendoarId(final String opendoarId) {
this.opendoarId = opendoarId;
}
public String getRorId() {
return this.rorId;
}
public void setRorId(final String rorId) {
this.rorId = rorId;
}
}

View File

@ -0,0 +1,184 @@
package eu.dnetlib.dhp.collection.plugin.base;
import static org.junit.jupiter.api.Assertions.assertEquals;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.HashSet;
import java.util.LinkedHashSet;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import java.util.Set;
import java.util.concurrent.atomic.AtomicInteger;
import org.apache.commons.io.IOUtils;
import org.apache.commons.lang3.StringUtils;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Encoders;
import org.apache.spark.sql.SparkSession;
import org.dom4j.Attribute;
import org.dom4j.Document;
import org.dom4j.DocumentException;
import org.dom4j.DocumentHelper;
import org.dom4j.Element;
import org.dom4j.Node;
import org.junit.jupiter.api.Disabled;
import org.junit.jupiter.api.Test;
import com.fasterxml.jackson.databind.ObjectMapper;
import eu.dnetlib.dhp.common.aggregation.AggregatorReport;
@Disabled
public class BaseCollectorIteratorTest {
@Test
void testImportFile() throws Exception {
long count = 0;
final BaseCollectorIterator iterator = new BaseCollectorIterator("base-sample.tar", new AggregatorReport());
final Map<String, Map<String, String>> collections = new HashMap<>();
final Map<String, AtomicInteger> fields = new HashMap<>();
final Set<String> types = new HashSet<>();
while (iterator.hasNext()) {
final Document record = DocumentHelper.parseText(iterator.next());
count++;
if ((count % 1000) == 0) {
System.out.println("# Read records: " + count);
}
// System.out.println(record.asXML());
for (final Object o : record.selectNodes("//*|//@*")) {
final String path = ((Node) o).getPath();
if (fields.containsKey(path)) {
fields.get(path).incrementAndGet();
} else {
fields.put(path, new AtomicInteger(1));
}
if (o instanceof Element) {
final Element n = (Element) o;
if ("collection".equals(n.getName())) {
final String collName = n.getText().trim();
if (StringUtils.isNotBlank(collName) && !collections.containsKey(collName)) {
final Map<String, String> collAttrs = new HashMap<>();
for (final Object ao : n.attributes()) {
collAttrs.put(((Attribute) ao).getName(), ((Attribute) ao).getValue());
}
collections.put(collName, collAttrs);
}
} else if ("type".equals(n.getName())) {
types.add(n.getText().trim());
}
}
}
}
final ObjectMapper mapper = new ObjectMapper();
for (final Entry<String, Map<String, String>> e : collections.entrySet()) {
System.out.println(e.getKey() + ": " + mapper.writeValueAsString(e.getValue()));
}
for (final Entry<String, AtomicInteger> e : fields.entrySet()) {
System.out.println(e.getKey() + ": " + e.getValue().get());
}
System.out.println("TYPES: ");
for (final String s : types) {
System.out.println(s);
}
assertEquals(30000, count);
}
@Test
public void testParquet() throws Exception {
final String xml = IOUtils.toString(getClass().getResourceAsStream("record.xml"));
final SparkSession spark = SparkSession.builder().master("local[*]").getOrCreate();
final List<BaseRecordInfo> ls = new ArrayList<>();
for (int i = 0; i < 10; i++) {
ls.add(extractInfo(xml));
}
final JavaRDD<BaseRecordInfo> rdd = JavaSparkContext
.fromSparkContext(spark.sparkContext())
.parallelize(ls);
final Dataset<BaseRecordInfo> df = spark
.createDataset(rdd.rdd(), Encoders.bean(BaseRecordInfo.class));
df.printSchema();
df.show(false);
}
private BaseRecordInfo extractInfo(final String s) {
try {
final Document record = DocumentHelper.parseText(s);
final BaseRecordInfo info = new BaseRecordInfo();
final Set<String> paths = new LinkedHashSet<>();
final Set<String> types = new LinkedHashSet<>();
final List<BaseCollectionInfo> colls = new ArrayList<>();
for (final Object o : record.selectNodes("//*|//@*")) {
paths.add(((Node) o).getPath());
if (o instanceof Element) {
final Element n = (Element) o;
final String nodeName = n.getName();
if ("collection".equals(nodeName)) {
final String collName = n.getText().trim();
if (StringUtils.isNotBlank(collName)) {
final BaseCollectionInfo coll = new BaseCollectionInfo();
coll.setId(collName);
coll.setOpendoarId(n.valueOf("@opendoar_id").trim());
coll.setRorId(n.valueOf("@ror_id").trim());
colls.add(coll);
}
} else if ("type".equals(nodeName)) {
types.add("TYPE: " + n.getText().trim());
} else if ("typenorm".equals(nodeName)) {
types.add("TYPE_NORM: " + n.getText().trim());
}
}
}
info.setId(record.valueOf("//*[local-name() = 'header']/*[local-name() = 'identifier']").trim());
info.getTypes().addAll(types);
info.getPaths().addAll(paths);
info.setCollections(colls);
return info;
} catch (final DocumentException e) {
throw new RuntimeException(e);
}
}
}

View File

@ -0,0 +1,32 @@
package eu.dnetlib.dhp.collection.plugin.base;
import static org.junit.jupiter.api.Assertions.assertFalse;
import static org.junit.jupiter.api.Assertions.assertTrue;
import java.util.Arrays;
import java.util.HashSet;
import java.util.Set;
import org.apache.commons.io.IOUtils;
import org.junit.jupiter.api.Test;
class BaseCollectorPluginTest {
@Test
void testFilterXml() throws Exception {
final String xml = IOUtils.toString(getClass().getResourceAsStream("record.xml"));
final Set<String> validIds = new HashSet<>(Arrays.asList("opendoar____::1234", "opendoar____::4567"));
final Set<String> validTypes = new HashSet<>(Arrays.asList("1", "121"));
final Set<String> validTypes2 = new HashSet<>(Arrays.asList("1", "11"));
assertTrue(BaseCollectorPlugin.filterXml(xml, validIds, validTypes));
assertTrue(BaseCollectorPlugin.filterXml(xml, validIds, new HashSet<>()));
assertFalse(BaseCollectorPlugin.filterXml(xml, new HashSet<>(), validTypes));
assertFalse(BaseCollectorPlugin.filterXml(xml, validIds, validTypes2));
}
}

View File

@ -0,0 +1,49 @@
package eu.dnetlib.dhp.collection.plugin.base;
import java.io.Serializable;
import java.util.ArrayList;
import java.util.List;
public class BaseRecordInfo implements Serializable {
private static final long serialVersionUID = -8848232018350074593L;
private String id;
private List<BaseCollectionInfo> collections = new ArrayList<>();
private List<String> paths = new ArrayList<>();
private List<String> types = new ArrayList<>();
public String getId() {
return this.id;
}
public void setId(final String id) {
this.id = id;
}
public List<String> getPaths() {
return this.paths;
}
public void setPaths(final List<String> paths) {
this.paths = paths;
}
public List<String> getTypes() {
return this.types;
}
public void setTypes(final List<String> types) {
this.types = types;
}
public List<BaseCollectionInfo> getCollections() {
return this.collections;
}
public void setCollections(final List<BaseCollectionInfo> collections) {
this.collections = collections;
}
}

View File

@ -0,0 +1,94 @@
package eu.dnetlib.dhp.collection.plugin.base;
import static org.junit.jupiter.api.Assertions.assertThrows;
import java.io.IOException;
import org.apache.commons.io.IOUtils;
import org.apache.spark.SparkConf;
import org.apache.spark.util.LongAccumulator;
import org.dom4j.io.SAXReader;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.Test;
import org.junit.jupiter.api.extension.ExtendWith;
import org.mockito.junit.jupiter.MockitoExtension;
import eu.dnetlib.dhp.aggregation.AbstractVocabularyTest;
import eu.dnetlib.dhp.aggregation.common.AggregationCounter;
import eu.dnetlib.dhp.schema.mdstore.MetadataRecord;
import eu.dnetlib.dhp.schema.mdstore.Provenance;
import eu.dnetlib.dhp.transformation.xslt.XSLTTransformationFunction;
import eu.dnetlib.enabling.is.lookup.rmi.ISLookUpException;
// @Disabled
@ExtendWith(MockitoExtension.class)
public class BaseTransfomationTest extends AbstractVocabularyTest {
private SparkConf sparkConf;
@BeforeEach
public void setUp() throws IOException, ISLookUpException {
setUpVocabulary();
this.sparkConf = new SparkConf();
this.sparkConf.setMaster("local[*]");
this.sparkConf.set("spark.driver.host", "localhost");
this.sparkConf.set("spark.ui.enabled", "false");
}
@Test
void testBase2ODF() throws Exception {
final MetadataRecord mr = new MetadataRecord();
mr.setProvenance(new Provenance("DSID", "DSNAME", "PREFIX"));
mr.setBody(IOUtils.toString(getClass().getResourceAsStream("record.xml")));
final XSLTTransformationFunction tr = loadTransformationRule("xml/base2odf.transformationRule.xml");
final MetadataRecord result = tr.call(mr);
System.out.println(result.getBody());
}
@Test
void testBase2OAF() throws Exception {
final MetadataRecord mr = new MetadataRecord();
mr.setProvenance(new Provenance("DSID", "DSNAME", "PREFIX"));
mr.setBody(IOUtils.toString(getClass().getResourceAsStream("record.xml")));
final XSLTTransformationFunction tr = loadTransformationRule("xml/base2oaf.transformationRule.xml");
final MetadataRecord result = tr.call(mr);
System.out.println(result.getBody());
}
@Test
void testBase2ODF_wrong_date() throws Exception {
final MetadataRecord mr = new MetadataRecord();
mr.setProvenance(new Provenance("DSID", "DSNAME", "PREFIX"));
mr.setBody(IOUtils.toString(getClass().getResourceAsStream("record_wrong_1.xml")));
final XSLTTransformationFunction tr = loadTransformationRule("xml/base2oaf.transformationRule.xml");
assertThrows(NullPointerException.class, () -> {
final MetadataRecord result = tr.call(mr);
System.out.println(result.getBody());
});
}
private XSLTTransformationFunction loadTransformationRule(final String path) throws Exception {
final String xslt = new SAXReader()
.read(this.getClass().getResourceAsStream(path))
.selectSingleNode("//CODE/*")
.asXML();
final LongAccumulator la = new LongAccumulator();
return new XSLTTransformationFunction(new AggregationCounter(la, la, la), xslt, 0, this.vocabularies);
}
}

View File

@ -9,6 +9,7 @@ import org.junit.jupiter.api.Test;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import eu.dnetlib.dhp.common.collection.CollectorException;
import eu.dnetlib.dhp.common.collection.HttpClientParams;
/**
@ -37,7 +38,7 @@ public class RestIteratorTest {
@Disabled
@Test
public void test() {
public void test() throws CollectorException {
HttpClientParams clientParams = new HttpClientParams();

View File

@ -0,0 +1,48 @@
package eu.dnetlib.dhp.collection.plugin.utils;
import static org.junit.jupiter.api.Assertions.assertEquals;
import org.junit.jupiter.api.Test;
class JsonUtilsTest {
static private String wrapped(String xml) {
return "<?xml version=\"1.0\" encoding=\"UTF-8\"?><recordWrap>" + xml + "</recordWrap>";
}
@Test
void keyStartWithDigit() {
assertEquals(
wrapped("<m_100><n_200v>null</n_200v></m_100>"),
JsonUtils.convertToXML("{\"100\" : {\"200v\" : null}}"));
}
@Test
void keyStartWithSpecialchars() {
assertEquals(
wrapped("<_parent><_nest1><_nest2>null</_nest2></_nest1></_parent>"),
JsonUtils.convertToXML("{\" parent\" : {\"-nest1\" : {\".nest2\" : null}}}"));
}
@Test
void encodeArray() {
assertEquals(
wrapped("<_parent.child>1</_parent.child><_parent.child>2</_parent.child>"),
JsonUtils.convertToXML("{\" parent.child\":[1, 2]}"));
}
@Test
void arrayOfObjects() {
assertEquals(
wrapped("<parent><id>1</id></parent><parent><id>2</id></parent>"),
JsonUtils.convertToXML("{\"parent\": [{\"id\": 1}, {\"id\": 2}]}"));
}
@Test
void removeControlCharacters() {
assertEquals(
wrapped("<m_100><n_200v>Test</n_200v></m_100>"),
JsonUtils.convertToXML("{\"100\" : {\"200v\" : \"\\u0000\\u000cTest\"}}"));
}
}

View File

@ -3,6 +3,7 @@ package eu.dnetlib.dhp.transformation;
import static eu.dnetlib.dhp.common.Constants.MDSTORE_DATA_PATH;
import static org.junit.jupiter.api.Assertions.assertEquals;
import static org.junit.jupiter.api.Assertions.assertThrows;
import java.io.IOException;
import java.nio.file.Path;
@ -279,6 +280,19 @@ class TransformationJobTest extends AbstractVocabularyTest {
// TODO Create significant Assert
}
@Test
public void testInvalidXSLT() throws Exception{
final MetadataRecord mr = new MetadataRecord();
mr.setProvenance(new Provenance("openaire____::cnr_explora", "CNR ExploRA", "cnr_________"));
mr.setBody(IOUtils.toString(getClass().getResourceAsStream("/eu/dnetlib/dhp/transform/input_cnr_explora.xml")));
// We Load the XSLT transformation Rule from the classpath
final XSLTTransformationFunction tr = loadTransformationRule("/eu/dnetlib/dhp/transform/invalid.xslt");
assertThrows(RuntimeException.class,()->tr.call(mr));
}
private XSLTTransformationFunction loadTransformationRule(final String path) throws Exception {
final String trValue = IOUtils.toString(this.getClass().getResourceAsStream(path));
final LongAccumulator la = new LongAccumulator();

View File

@ -0,0 +1,19 @@
DOI,OAID,level1,level2,level3,level4,score_for_L3,score_for_L4
N/A,78975075580c::e680668c98366c9cd6349afc62486a7f,03 medical and health sciences,0301 basic medicine,030104 developmental biology,N/A,0.5,0.0
N/A,78975075580c::e680668c98366c9cd6349afc62486a7f,03 medical and health sciences,0303 health sciences,030304 developmental biology,N/A,0.5,0.0
N/A,od______2806::a1da9d2678b12969a9ab5f50b5e71d0a,05 social sciences,0501 psychology and cognitive sciences,050109 social psychology,05010904 Group processes/Collective identity,0.5589094161987305,0.5166763067245483
N/A,od______2806::a1da9d2678b12969a9ab5f50b5e71d0a,05 social sciences,0501 psychology and cognitive sciences,050105 experimental psychology,05010501 Emotion/Affective science,0.44109055399894714,0.4833236634731293
N/A,doajarticles::76535d77fd2a5fe9810aefafffb8ef6c,05 social sciences,0502 economics and business,050203 business & management,05020302 Supply chain management/Business terms,0.5459638833999634,0.5460261106491089
N/A,doajarticles::76535d77fd2a5fe9810aefafffb8ef6c,05 social sciences,0502 economics and business,050211 marketing,05021102 Services marketing/Retailing,0.4540362060070038,0.4539738595485687
N/A,od_______156::a3a0119c6d9d3a66943f8da042e97a5e,01 natural sciences,0105 earth and related environmental sciences,010504 meteorology & atmospheric sciences,01050407 Geomagnetism/Ionosphere,0.5131047964096069,0.4990350902080536
N/A,od_______156::a3a0119c6d9d3a66943f8da042e97a5e,01 natural sciences,0105 earth and related environmental sciences,010502 geochemistry & geophysics,01050203 Seismology/Seismology measurement,0.4868951737880707,0.500964879989624
N/A,od______2806::4b9a664dd6b8b04204cb613e7bc9c873,03 medical and health sciences,0302 clinical medicine,030220 oncology & carcinogenesis,03022002 Medical imaging/Medical physics,0.5068133473396301,0.10231181626910052
N/A,od______2806::4b9a664dd6b8b04204cb613e7bc9c873,03 medical and health sciences,0302 clinical medicine,030204 cardiovascular system & hematology,N/A,0.49318668246269226,0.0
N/A,od______3341::ef754de29464abf9bc9b99664630ce74,03 medical and health sciences,0302 clinical medicine,030220 oncology & carcinogenesis,03022012 Oncology/Infectious causes of cancer,0.5,0.5
N/A,od______3341::ef754de29464abf9bc9b99664630ce74,03 medical and health sciences,0302 clinical medicine,030220 oncology & carcinogenesis,03022012 Oncology/Infectious causes of cancer,0.5,0.5
N/A,od______3978::6704dcced0fe3dd6fbf985dc2507f61c,03 medical and health sciences,0302 clinical medicine,030217 neurology & neurosurgery,03021702 Aging-associated diseases/Cognitive disorders,0.5134317874908447,0.09614889098529535
N/A,od______3978::6704dcced0fe3dd6fbf985dc2507f61c,03 medical and health sciences,0301 basic medicine,030104 developmental biology,N/A,0.48656824231147766,0.0
N/A,dedup_wf_001::b77264819800b90c0328c4d17eea5c1a,02 engineering and technology,0209 industrial biotechnology,020901 industrial engineering & automation,02090105 Control theory/Advanced driver assistance systems,0.5178514122962952,0.5198937654495239
N/A,dedup_wf_001::b77264819800b90c0328c4d17eea5c1a,02 engineering and technology,"0202 electrical engineering, electronic engineering, information engineering",020201 artificial intelligence & image processing,02020108 Fuzzy logic/Artificial neural networks/Computational neuroscience,0.48214852809906006,0.4801062345504761
N/A,od______2806::a938609e9f36ada6629a1bcc50c88230,03 medical and health sciences,0302 clinical medicine,030217 neurology & neurosurgery,03021708 Neurotrauma/Stroke,0.5014800429344177,0.5109656453132629
N/A,od______2806::a938609e9f36ada6629a1bcc50c88230,02 engineering and technology,0206 medical engineering,020601 biomedical engineering,02060102 Medical terminology/Patient,0.4985199570655823,0.4890343248844147
1 DOI OAID level1 level2 level3 level4 score_for_L3 score_for_L4
2 N/A 78975075580c::e680668c98366c9cd6349afc62486a7f 03 medical and health sciences 0301 basic medicine 030104 developmental biology N/A 0.5 0.0
3 N/A 78975075580c::e680668c98366c9cd6349afc62486a7f 03 medical and health sciences 0303 health sciences 030304 developmental biology N/A 0.5 0.0
4 N/A od______2806::a1da9d2678b12969a9ab5f50b5e71d0a 05 social sciences 0501 psychology and cognitive sciences 050109 social psychology 05010904 Group processes/Collective identity 0.5589094161987305 0.5166763067245483
5 N/A od______2806::a1da9d2678b12969a9ab5f50b5e71d0a 05 social sciences 0501 psychology and cognitive sciences 050105 experimental psychology 05010501 Emotion/Affective science 0.44109055399894714 0.4833236634731293
6 N/A doajarticles::76535d77fd2a5fe9810aefafffb8ef6c 05 social sciences 0502 economics and business 050203 business & management 05020302 Supply chain management/Business terms 0.5459638833999634 0.5460261106491089
7 N/A doajarticles::76535d77fd2a5fe9810aefafffb8ef6c 05 social sciences 0502 economics and business 050211 marketing 05021102 Services marketing/Retailing 0.4540362060070038 0.4539738595485687
8 N/A od_______156::a3a0119c6d9d3a66943f8da042e97a5e 01 natural sciences 0105 earth and related environmental sciences 010504 meteorology & atmospheric sciences 01050407 Geomagnetism/Ionosphere 0.5131047964096069 0.4990350902080536
9 N/A od_______156::a3a0119c6d9d3a66943f8da042e97a5e 01 natural sciences 0105 earth and related environmental sciences 010502 geochemistry & geophysics 01050203 Seismology/Seismology measurement 0.4868951737880707 0.500964879989624
10 N/A od______2806::4b9a664dd6b8b04204cb613e7bc9c873 03 medical and health sciences 0302 clinical medicine 030220 oncology & carcinogenesis 03022002 Medical imaging/Medical physics 0.5068133473396301 0.10231181626910052
11 N/A od______2806::4b9a664dd6b8b04204cb613e7bc9c873 03 medical and health sciences 0302 clinical medicine 030204 cardiovascular system & hematology N/A 0.49318668246269226 0.0
12 N/A od______3341::ef754de29464abf9bc9b99664630ce74 03 medical and health sciences 0302 clinical medicine 030220 oncology & carcinogenesis 03022012 Oncology/Infectious causes of cancer 0.5 0.5
13 N/A od______3341::ef754de29464abf9bc9b99664630ce74 03 medical and health sciences 0302 clinical medicine 030220 oncology & carcinogenesis 03022012 Oncology/Infectious causes of cancer 0.5 0.5
14 N/A od______3978::6704dcced0fe3dd6fbf985dc2507f61c 03 medical and health sciences 0302 clinical medicine 030217 neurology & neurosurgery 03021702 Aging-associated diseases/Cognitive disorders 0.5134317874908447 0.09614889098529535
15 N/A od______3978::6704dcced0fe3dd6fbf985dc2507f61c 03 medical and health sciences 0301 basic medicine 030104 developmental biology N/A 0.48656824231147766 0.0
16 N/A dedup_wf_001::b77264819800b90c0328c4d17eea5c1a 02 engineering and technology 0209 industrial biotechnology 020901 industrial engineering & automation 02090105 Control theory/Advanced driver assistance systems 0.5178514122962952 0.5198937654495239
17 N/A dedup_wf_001::b77264819800b90c0328c4d17eea5c1a 02 engineering and technology 0202 electrical engineering, electronic engineering, information engineering 020201 artificial intelligence & image processing 02020108 Fuzzy logic/Artificial neural networks/Computational neuroscience 0.48214852809906006 0.4801062345504761
18 N/A od______2806::a938609e9f36ada6629a1bcc50c88230 03 medical and health sciences 0302 clinical medicine 030217 neurology & neurosurgery 03021708 Neurotrauma/Stroke 0.5014800429344177 0.5109656453132629
19 N/A od______2806::a938609e9f36ada6629a1bcc50c88230 02 engineering and technology 0206 medical engineering 020601 biomedical engineering 02060102 Medical terminology/Patient 0.4985199570655823 0.4890343248844147

View File

@ -0,0 +1,18 @@
{"doi":"n/a","oaid":"od______3341::ef754de29464abf9bc9b99664630ce74","level1":"03 medical and health sciences","level2":"0302 clinical medicine","level3":"030220 oncology & carcinogenesis","level4":"03022012 Oncology/Infectious causes of cancer","scoreL3":"0.5","scoreL4":"0.5"}
{"doi":"n/a","oaid":"78975075580c::e680668c98366c9cd6349afc62486a7f","level1":"03 medical and health sciences","level2":"0301 basic medicine","level3":"030104 developmental biology","level4":"N/A","scoreL3":"0.5","scoreL4":"0.0"}
{"doi":"n/a","oaid":"od______3341::ef754de29464abf9bc9b99664630ce74","level1":"03 medical and health sciences","level2":"0302 clinical medicine","level3":"030220 oncology & carcinogenesis","level4":"03022012 Oncology/Infectious causes of cancer","scoreL3":"0.5","scoreL4":"0.5"}
{"doi":"n/a","oaid":"78975075580c::e680668c98366c9cd6349afc62486a7f","level1":"03 medical and health sciences","level2":"0303 health sciences","level3":"030304 developmental biology","level4":"N/A","scoreL3":"0.5","scoreL4":"0.0"}
{"doi":"n/a","oaid":"od______3978::6704dcced0fe3dd6fbf985dc2507f61c","level1":"03 medical and health sciences","level2":"0302 clinical medicine","level3":"030217 neurology & neurosurgery","level4":"03021702 Aging-associated diseases/Cognitive disorders","scoreL3":"0.5134317874908447","scoreL4":"0.09614889098529535"}
{"doi":"n/a","oaid":"od______2806::a1da9d2678b12969a9ab5f50b5e71d0a","level1":"05 social sciences","level2":"0501 psychology and cognitive sciences","level3":"050109 social psychology","level4":"05010904 Group processes/Collective identity","scoreL3":"0.5589094161987305","scoreL4":"0.5166763067245483"}
{"doi":"n/a","oaid":"od______3978::6704dcced0fe3dd6fbf985dc2507f61c","level1":"03 medical and health sciences","level2":"0301 basic medicine","level3":"030104 developmental biology","level4":"N/A","scoreL3":"0.48656824231147766","scoreL4":"0.0"}
{"doi":"n/a","oaid":"od______2806::a1da9d2678b12969a9ab5f50b5e71d0a","level1":"05 social sciences","level2":"0501 psychology and cognitive sciences","level3":"050105 experimental psychology","level4":"05010501 Emotion/Affective science","scoreL3":"0.44109055399894714","scoreL4":"0.4833236634731293"}
{"doi":"n/a","oaid":"dedup_wf_001::b77264819800b90c0328c4d17eea5c1a","level1":"02 engineering and technology","level2":"0209 industrial biotechnology","level3":"020901 industrial engineering & automation","level4":"02090105 Control theory/Advanced driver assistance systems","scoreL3":"0.5178514122962952","scoreL4":"0.5198937654495239"}
{"doi":"n/a","oaid":"doajarticles::76535d77fd2a5fe9810aefafffb8ef6c","level1":"05 social sciences","level2":"0502 economics and business","level3":"050203 business & management","level4":"05020302 Supply chain management/Business terms","scoreL3":"0.5459638833999634","scoreL4":"0.5460261106491089"}
{"doi":"n/a","oaid":"doajarticles::76535d77fd2a5fe9810aefafffb8ef6c","level1":"05 social sciences","level2":"0502 economics and business","level3":"050211 marketing","level4":"05021102 Services marketing/Retailing","scoreL3":"0.4540362060070038","scoreL4":"0.4539738595485687"}
{"doi":"n/a","oaid":"dedup_wf_001::b77264819800b90c0328c4d17eea5c1a","level1":"02 engineering and technology","level2":"0202 electrical engineering, electronic engineering, information engineering","level3":"020201 artificial intelligence & image processing","level4":"02020108 Fuzzy logic/Artificial neural networks/Computational neuroscience","scoreL3":"0.48214852809906006","scoreL4":"0.4801062345504761"}
{"doi":"n/a","oaid":"od_______156::a3a0119c6d9d3a66943f8da042e97a5e","level1":"01 natural sciences","level2":"0105 earth and related environmental sciences","level3":"010504 meteorology & atmospheric sciences","level4":"01050407 Geomagnetism/Ionosphere","scoreL3":"0.5131047964096069","scoreL4":"0.4990350902080536"}
{"doi":"n/a","oaid":"od______2806::a938609e9f36ada6629a1bcc50c88230","level1":"03 medical and health sciences","level2":"0302 clinical medicine","level3":"030217 neurology & neurosurgery","level4":"03021708 Neurotrauma/Stroke","scoreL3":"0.5014800429344177","scoreL4":"0.5109656453132629"}
{"doi":"n/a","oaid":"od_______156::a3a0119c6d9d3a66943f8da042e97a5e","level1":"01 natural sciences","level2":"0105 earth and related environmental sciences","level3":"010502 geochemistry & geophysics","level4":"01050203 Seismology/Seismology measurement","scoreL3":"0.4868951737880707","scoreL4":"0.500964879989624"}
{"doi":"n/a","oaid":"od______2806::a938609e9f36ada6629a1bcc50c88230","level1":"02 engineering and technology","level2":"0206 medical engineering","level3":"020601 biomedical engineering","level4":"02060102 Medical terminology/Patient","scoreL3":"0.4985199570655823","scoreL4":"0.4890343248844147"}
{"doi":"n/a","oaid":"od______2806::4b9a664dd6b8b04204cb613e7bc9c873","level1":"03 medical and health sciences","level2":"0302 clinical medicine","level3":"030220 oncology & carcinogenesis","level4":"03022002 Medical imaging/Medical physics","scoreL3":"0.5068133473396301","scoreL4":"0.10231181626910052"}
{"doi":"n/a","oaid":"od______2806::4b9a664dd6b8b04204cb613e7bc9c873","level1":"03 medical and health sciences","level2":"0302 clinical medicine","level3":"030204 cardiovascular system & hematology","level4":"N/A","scoreL3":"0.49318668246269226","scoreL4":"0.0"}

File diff suppressed because one or more lines are too long

View File

@ -0,0 +1,31 @@
{"cited":"br/061201599020", "citing":"br/06203041400","oci":"oci:06701327944-06504326071"}
{"cited":"br/061201599020","citing":"br/06502272390","oci":"oci:06502272390-061301355525"}
{"cited":"br/061201599020", "citing":"br/06120941789","oci":"oci:0670804699-067055659"}
{"cited":"br/06210273177","citing":"br/06203041400","oci":"oci:061502003994-062201281456"}
{"cited":"br/06210273177", "citing":"br/06502272390","oci":"oci:06502272390-0660806688"}
{"cited":"br/06210273177", "citing":"br/06120941789","oci":"oci:06502307119-0620223645"}
{"cited":"br/0660613430","citing":"br/06203041400","oci":"oci:061502004011-061902692285"}
{"cited":"br/0660613430", "citing":"br/06502272390","oci":"oci:0660549063-0610398792"}
{"cited":"br/0660613430", "citing":"br/06120941789","oci":"oci:06420189324-06301543046"}
{"cited":"br/062602732073","citing":"br/06203041400","oci":"oci:06380130275-061502004367"}
{"cited":"br/062602732073","citing":"br/06502272390","oci":"oci:062403449086-062501448395"}
{"cited":"br/062602732073","citing":"br/06120941789","oci":"oci:06420189328-061202007182"}
{"cited":"br/061103703697","citing":"br/06203041400","oci":"oci:062603906965-061701362658"}
{"cited":"br/061103703697", "citing":"br/06502272390","oci":"oci:0670294309-06104327031"}
{"cited":"br/061103703697","citing":"br/06120941789","oci":"oci:061702060228-061301712529"}
{"cited":"br/06230199640", "citing":"br/0670517081","oci":"oci:06901104174-06503692526"}
{"cited":"br/061703513967","citing":"br/061702310822","oci":"oci:061702310822-061703513967"}
{"cited":"br/062104002953","citing":"br/061702311472","oci":"oci:061702311472-062104002953"}
{"cited":"br/061101204417","citing":"br/062102701590","oci":"oci:062102701590-061101204417"}
{"cited":"br/062403787088","citing":"br/061401499173","oci":"oci:061401499173-062403787088"}
{"cited":"br/061203576338","citing":"br/06110279619","oci":"oci:06110279619-061203576338"}
{"cited":"br/061601962207","citing":"br/061502004018","oci":"oci:061502004018-061601962207"}
{"cited":"br/06101014588", "citing":"br/061502004027","oci":"oci:061502004027-06101014588"}
{"cited":"br/06704040804", "citing":"br/06220799044","oci":"oci:06220799044-06704040804"}
{"cited":"br/061401105151","citing":"br/061502004037","oci":"oci:061502004037-061401105151"}
{"cited":"br/0640821079", "citing":"br/061702311537","oci":"oci:061702311537-0640821079"}
{"cited":"br/06604165310", "citing":"br/062501970289","oci":"oci:062501970289-06604165310"}
{"cited":"br/061501351689","citing":"br/061203895786","oci":"oci:061203895786-061501351689"}
{"cited":"br/06202223692", "citing":"br/06110298832","oci":"oci:06110298832-06202223692"}
{"cited":"br/06104310727", "citing":"br/0660439086","oci":"oci:0660439086-06104310727"}
{"cited":"br/06150216214", "citing":"br/06340150329","oci":"oci:06340150329-06150216214"}

View File

@ -0,0 +1,48 @@
omid,id
br/061201599020,doi:10.1142/s0219887817501687
br/06203041400,doi:10.1111/j.1523-5378.2005.00327.x pmid:16104945
br/06210273177,doi:10.1090/qam/20394
br/06502272390,pmid:32235596 doi:10.3390/nano10040644
br/0660613430,doi:10.1007/bf00470411
br/06120941789,doi:10.1098/rspa.2006.1747
br/062602732073,doi:10.1007/978-3-642-38844-6_25
br/06230199640,pmid:25088780 doi:10.1016/j.ymeth.2014.07.008
br/061103703697,pmid:2682767
br/0670517081,doi:10.1016/j.foodpol.2021.102189
br/06502310477,doi:10.1142/s0218127416500450
br/06520113284,doi:10.1109/cfasta57821.2023.10243367
br/062303652439,pmid:5962654 doi:10.1016/0020-708x(66)90001-9
br/06250691436,doi:10.1042/bst20150052 pmid:26009172
br/061201665577,doi:10.1097/00115550-200205000-00018
br/06503490336,pmid:34689254 doi:10.1007/s10072-021-05687-0
br/06220615942,pmid:25626134 doi:10.1016/j.jcis.2015.01.008
br/061103389243,doi:10.4324/9780203702819-10
br/062303011271,doi:10.1109/icassp.2011.5946250
br/061302926083,doi:10.4018/978-1-6684-3937-1.ch002
br/061402485360,doi:10.1109/iciict.2015.7396079
br/06410101083,doi:10.1016/j.autcon.2023.104828
br/062202243386,doi:10.1016/0001-8791(81)90022-1
br/06170421486,doi:10.1130/0016-7606(2003)115<0166:dsagmf>2.0.co;2
br/061201983865,doi:10.4324/9781315109008 isbn:9781315109008
br/061701697230,doi:10.1016/j.trd.2012.07.006
br/061201137111,doi:10.1109/access.2020.2971656
br/06120436283,pmid:2254430 doi:10.1128/jcm.28.11.2551-2554.1990
br/061903968916,doi:10.1111/j.1742-1241.1988.tb08627.x
br/06201583482,doi:10.1016/0016-5085(78)93139-6
br/06130338317,doi:10.2134/agronj1952.00021962004400080013x
br/062601538320,doi:10.1371/journal.pone.0270593 pmid:35789338
br/062401098626,pmid:22385804 doi:10.1016/j.talanta.2011.12.034
br/06190436492,doi:10.1039/c7dt01499f pmid:28644489
br/06202819247,doi:10.1007/978-3-319-45823-6_57
br/0648013560,doi:10.1080/14772000.2012.705356
br/0690214059,doi:10.2752/175630608x329217
br/06601640415,doi:10.1080/18128600508685647
br/061503394761,doi:10.1002/0471443395.img018
br/061702861849,pmid:31203682 doi:10.1080/10428194.2019.1627538
br/06450133713,doi:10.1093/acprof:oso/9780199670888.003.0008
br/0628074892,doi:10.1097/hnp.0000000000000597
br/061601032219,doi:10.1002/bdm.2102
br/06602079930,doi:10.1101/2020.08.25.267500
br/0604192147,doi:10.11501/3307395
br/061101933800,doi:10.1142/s0217732398002242
br/06504184118,pmid:10091417
1 omid id
2 br/061201599020 doi:10.1142/s0219887817501687
3 br/06203041400 doi:10.1111/j.1523-5378.2005.00327.x pmid:16104945
4 br/06210273177 doi:10.1090/qam/20394
5 br/06502272390 pmid:32235596 doi:10.3390/nano10040644
6 br/0660613430 doi:10.1007/bf00470411
7 br/06120941789 doi:10.1098/rspa.2006.1747
8 br/062602732073 doi:10.1007/978-3-642-38844-6_25
9 br/06230199640 pmid:25088780 doi:10.1016/j.ymeth.2014.07.008
10 br/061103703697 pmid:2682767
11 br/0670517081 doi:10.1016/j.foodpol.2021.102189
12 br/06502310477 doi:10.1142/s0218127416500450
13 br/06520113284 doi:10.1109/cfasta57821.2023.10243367
14 br/062303652439 pmid:5962654 doi:10.1016/0020-708x(66)90001-9
15 br/06250691436 doi:10.1042/bst20150052 pmid:26009172
16 br/061201665577 doi:10.1097/00115550-200205000-00018
17 br/06503490336 pmid:34689254 doi:10.1007/s10072-021-05687-0
18 br/06220615942 pmid:25626134 doi:10.1016/j.jcis.2015.01.008
19 br/061103389243 doi:10.4324/9780203702819-10
20 br/062303011271 doi:10.1109/icassp.2011.5946250
21 br/061302926083 doi:10.4018/978-1-6684-3937-1.ch002
22 br/061402485360 doi:10.1109/iciict.2015.7396079
23 br/06410101083 doi:10.1016/j.autcon.2023.104828
24 br/062202243386 doi:10.1016/0001-8791(81)90022-1
25 br/06170421486 doi:10.1130/0016-7606(2003)115<0166:dsagmf>2.0.co;2
26 br/061201983865 doi:10.4324/9781315109008 isbn:9781315109008
27 br/061701697230 doi:10.1016/j.trd.2012.07.006
28 br/061201137111 doi:10.1109/access.2020.2971656
29 br/06120436283 pmid:2254430 doi:10.1128/jcm.28.11.2551-2554.1990
30 br/061903968916 doi:10.1111/j.1742-1241.1988.tb08627.x
31 br/06201583482 doi:10.1016/0016-5085(78)93139-6
32 br/06130338317 doi:10.2134/agronj1952.00021962004400080013x
33 br/062601538320 doi:10.1371/journal.pone.0270593 pmid:35789338
34 br/062401098626 pmid:22385804 doi:10.1016/j.talanta.2011.12.034
35 br/06190436492 doi:10.1039/c7dt01499f pmid:28644489
36 br/06202819247 doi:10.1007/978-3-319-45823-6_57
37 br/0648013560 doi:10.1080/14772000.2012.705356
38 br/0690214059 doi:10.2752/175630608x329217
39 br/06601640415 doi:10.1080/18128600508685647
40 br/061503394761 doi:10.1002/0471443395.img018
41 br/061702861849 pmid:31203682 doi:10.1080/10428194.2019.1627538
42 br/06450133713 doi:10.1093/acprof:oso/9780199670888.003.0008
43 br/0628074892 doi:10.1097/hnp.0000000000000597
44 br/061601032219 doi:10.1002/bdm.2102
45 br/06602079930 doi:10.1101/2020.08.25.267500
46 br/0604192147 doi:10.11501/3307395
47 br/061101933800 doi:10.1142/s0217732398002242
48 br/06504184118 pmid:10091417

View File

@ -0,0 +1,27 @@
{"oci":"oci:06701327944-06504326071","citing":"16104945","citing_pid":"pmid","cited":"10.1142/s0219887817501687","cited_pid":"doi"}
{"oci":"oci:06701327944-06504326071","citing":"10.1111/j.1523-5378.2005.00327.x","citing_pid":"doi","cited":"10.1142/s0219887817501687","cited_pid":"doi"}
{"oci":"oci:06502272390-061301355525","citing":"10.3390/nano10040644","citing_pid":"doi","cited":"10.1142/s0219887817501687","cited_pid":"doi"}
{"oci":"oci:06502272390-061301355525","citing":"32235596","citing_pid":"pmid","cited":"10.1142/s0219887817501687","cited_pid":"doi"}
{"oci":"oci:0670804699-067055659","citing":"10.1098/rspa.2006.1747","citing_pid":"doi","cited":"10.1142/s0219887817501687","cited_pid":"doi"}
{"oci":"oci:061502003994-062201281456","citing":"16104945","citing_pid":"pmid","cited":"10.1090/qam/20394","cited_pid":"doi"}
{"oci":"oci:061502003994-062201281456","citing":"10.1111/j.1523-5378.2005.00327.x","citing_pid":"doi","cited":"10.1090/qam/20394","cited_pid":"doi"}
{"oci":"oci:06502272390-0660806688","citing":"10.3390/nano10040644","citing_pid":"doi","cited":"10.1090/qam/20394","cited_pid":"doi"}
{"oci":"oci:06502272390-0660806688","citing":"32235596","citing_pid":"pmid","cited":"10.1090/qam/20394","cited_pid":"doi"}
{"oci":"oci:06502307119-0620223645","citing":"10.1098/rspa.2006.1747","citing_pid":"doi","cited":"10.1090/qam/20394","cited_pid":"doi"}
{"oci":"oci:061502004011-061902692285","citing":"16104945","citing_pid":"pmid","cited":"10.1007/bf00470411","cited_pid":"doi"}
{"oci":"oci:061502004011-061902692285","citing":"10.1111/j.1523-5378.2005.00327.x","citing_pid":"doi","cited":"10.1007/bf00470411","cited_pid":"doi"}
{"oci":"oci:0660549063-0610398792","citing":"10.3390/nano10040644","citing_pid":"doi","cited":"10.1007/bf00470411","cited_pid":"doi"}
{"oci":"oci:0660549063-0610398792","citing":"32235596","citing_pid":"pmid","cited":"10.1007/bf00470411","cited_pid":"doi"}
{"oci":"oci:06420189324-06301543046","citing":"10.1098/rspa.2006.1747","citing_pid":"doi","cited":"10.1007/bf00470411","cited_pid":"doi"}
{"oci":"oci:06380130275-061502004367","citing":"16104945","citing_pid":"pmid","cited":"10.1007/978-3-642-38844-6_25","cited_pid":"doi"}
{"oci":"oci:06380130275-061502004367","citing":"10.1111/j.1523-5378.2005.00327.x","citing_pid":"doi","cited":"10.1007/978-3-642-38844-6_25","cited_pid":"doi"}
{"oci":"oci:062403449086-062501448395","citing":"10.3390/nano10040644","citing_pid":"doi","cited":"10.1007/978-3-642-38844-6_25","cited_pid":"doi"}
{"oci":"oci:062403449086-062501448395","citing":"32235596","citing_pid":"pmid","cited":"10.1007/978-3-642-38844-6_25","cited_pid":"doi"}
{"oci":"oci:06420189328-061202007182","citing":"10.1098/rspa.2006.1747","citing_pid":"doi","cited":"10.1007/978-3-642-38844-6_25","cited_pid":"doi"}
{"oci":"oci:062603906965-061701362658","citing":"16104945","citing_pid":"pmid","cited":"2682767","cited_pid":"pmid"}
{"oci":"oci:062603906965-061701362658","citing":"10.1111/j.1523-5378.2005.00327.x","citing_pid":"doi","cited":"2682767","cited_pid":"pmid"}
{"oci":"oci:0670294309-06104327031","citing":"10.3390/nano10040644","citing_pid":"doi","cited":"2682767","cited_pid":"pmid"}
{"oci":"oci:0670294309-06104327031","citing":"32235596","citing_pid":"pmid","cited":"2682767","cited_pid":"pmid"}
{"oci":"oci:061702060228-061301712529","citing":"10.1098/rspa.2006.1747","citing_pid":"doi","cited":"2682767","cited_pid":"pmid"}
{"oci":"oci:06901104174-06503692526","citing":"10.1016/j.foodpol.2021.102189","citing_pid":"doi","cited":"10.1016/j.ymeth.2014.07.008","cited_pid":"doi"}
{"oci":"oci:06901104174-06503692526","citing":"10.1016/j.foodpol.2021.102189","citing_pid":"doi","cited":"25088780","cited_pid":"pmid"}

View File

@ -0,0 +1,58 @@
<record>
<header xmlns="http://www.openarchives.org/OAI/2.0/">
<identifier>ftdoajarticles:oai:doaj.org/article:e2d5b5126b2d4e479933cc7f9a9ae0c1</identifier>
<datestamp>2022-12-31T11:48:55Z</datestamp>
</header>
<metadata xmlns="http://www.openarchives.org/OAI/2.0/" xmlns:base_dc="http://oai.base-search.net/base_dc/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:dc="http://purl.org/dc/elements/1.1/">
<base_dc:dc xsi:schemaLocation="http://oai.base-search.net/base_dc/ http://oai.base-search.net/base_dc/base_dc.xsd">
<base_dc:global_id>ftdoajarticles:oai:doaj.org/article:e2d5b5126b2d4e479933cc7f9a9ae0c1</base_dc:global_id>
<base_dc:continent>cww</base_dc:continent>
<base_dc:country>org</base_dc:country>
<base_dc:collection opendoar_id="1234" ror_id="ror1234">ftdoajarticles</base_dc:collection>
<base_dc:collname>TEST REPO</base_dc:collname>
<dc:title>Assessment of cultural heritage: the legislative and methodological framework of Russian Federation</dc:title>
<dc:creator>ALBU, Svetlana</dc:creator>
<dc:creator>LEȘAN, Anna</dc:creator>
<dc:subject>architectural heritage</dc:subject>
<dc:subject>evaluation of architectural heritage</dc:subject>
<dc:subject>types of values</dc:subject>
<dc:subject>experience of russian federation</dc:subject>
<dc:subject>Social Sciences</dc:subject>
<dc:subject>H</dc:subject>
<dc:description>Architectural heritage is the real estate inheritance by population of a country becoming an extremely valuable and specific category, preserving and capitalizing on those assets requires considerable effort. The state does not have sufficient means to maintain and preserve cultural heritage, as a result it is included in the civil circuit. The transfer of property right or of some partial rights over the architectural patrimony is accompanied by the necessity to estimate the value of goods. In this article, the authors examine the experience of Russian Federation (one of the largest countries with a huge architectural heritage) on the legislative framework of architectural and methodological heritage of architectural heritage assessment. The particularities of cultural assets valuation compared to other categories of real estate are examined, as well as the methodological aspects (types of values, methods applied in valuation, approaches according to the purpose of valuation) regarding the valuation of real estate with architectural value in Russian Federation.</dc:description>
<dc:publisher>Technical University of Moldova</dc:publisher>
<dc:date>2020-09-01T00:00:00Z</dc:date>
<base_dc:year>2020</base_dc:year>
<dc:type>article</dc:type>
<base_dc:typenorm>121</base_dc:typenorm>
<dc:identifier>https://doi.org/10.5281/zenodo.3971988</dc:identifier>
<dc:identifier>https://doaj.org/article/e2d5b5126b2d4e479933cc7f9a9ae0c1</dc:identifier>
<base_dc:link>https://doi.org/10.5281/zenodo.3971988</base_dc:link>
<dc:source>Journal of Social Sciences, Vol 3, Iss 3, Pp 134-143 (2020)</dc:source>
<dc:language>EN</dc:language>
<dc:language>FR</dc:language>
<dc:language>RO</dc:language>
<dc:relation>http://ibn.idsi.md/sites/default/files/imag_file/JSS-3-2020_134-143.pdf</dc:relation>
<dc:relation>https://doaj.org/toc/2587-3490</dc:relation>
<dc:relation>https://doaj.org/toc/2587-3504</dc:relation>
<dc:relation>doi:10.5281/zenodo.3971988</dc:relation>
<dc:relation>2587-3490</dc:relation>
<dc:relation>2587-3504</dc:relation>
<dc:relation>https://doaj.org/article/e2d5b5126b2d4e479933cc7f9a9ae0c1</dc:relation>
<base_dc:autoclasscode type="ddc">720</base_dc:autoclasscode>
<base_dc:authod_id>
<base_dc:creator_name>ALBU, Svetlana</base_dc:creator_name>
<base_dc:creator_id>https://orcid.org/0000-0002-8648-950X</base_dc:creator_id>
</base_dc:authod_id>
<base_dc:authod_id>
<base_dc:creator_name>LEȘAN, Anna</base_dc:creator_name>
<base_dc:creator_id>https://orcid.org/0000-0003-3284-0525</base_dc:creator_id>
</base_dc:authod_id>
<base_dc:doi>https://doi.org/10.5281/zenodo.3971988</base_dc:doi>
<base_dc:oa>1</base_dc:oa>
<base_dc:lang>eng</base_dc:lang>
<base_dc:lang>fre</base_dc:lang>
<base_dc:lang>rum</base_dc:lang>
</base_dc:dc>
</metadata>
</record>

View File

@ -0,0 +1,36 @@
<?xml version="1.0" encoding="UTF-8"?>
<record>
<header xmlns="http://www.openarchives.org/OAI/2.0/">
<identifier>ftunivminnesdc:oai:conservancy.umn.edu:11299/109914</identifier>
<datestamp>2023-07-18T20:05:40Z</datestamp>
</header>
<metadata xmlns="http://www.openarchives.org/OAI/2.0/" xmlns:base_dc="http://oai.base-search.net/base_dc/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:dc="http://purl.org/dc/elements/1.1/">
<base_dc:dc xsi:schemaLocation="http://oai.base-search.net/base_dc/ http://oai.base-search.net/base_dc/base_dc.xsd">
<base_dc:global_id>ftunivminnesdc:oai:conservancy.umn.edu:11299/109914</base_dc:global_id>
<base_dc:continent>cna</base_dc:continent>
<base_dc:country>us</base_dc:country>
<base_dc:collection opendoar_id="1008">ftunivminnesdc</base_dc:collection>
<base_dc:collname>University of Minnesota Digital Conservancy</base_dc:collname>
<dc:title>An Experimental Investigation of the influence of an air bubble layer on radiated noise and surface pressure fluctuations in a turbulent boundary layer</dc:title>
<dc:creator>Killen, John M.</dc:creator>
<dc:subject>Boundary layer noise</dc:subject>
<dc:subject>Kinetic energy dissipation</dc:subject>
<dc:subject>Air bubble layer</dc:subject>
<dc:subject>Maximum stable bubble size</dc:subject>
<dc:subject>bubble noise spectra</dc:subject>
<dc:description>between the noise spectra of a layer of air bubbles in a turbulent flow and' a maximum stable bubble size which can exist in the same flow. An air bubbie layer with individual bubble sizes greater than a maximum stable size was introduced into the boundary layer of water flowing along a smooth flat plate. It was found that the maximum stable bubble size was related to the turbulent kinetic energy dissipation as had been previously shown by other investigators ; Office of Naval Research Arlington, Virginia 22217</dc:description>
<dc:date>1981-09-31T21:42:52Z</dc:date>
<base_dc:year>1981</base_dc:year>
<dc:type>Report</dc:type>
<base_dc:typenorm>14</base_dc:typenorm>
<dc:identifier>http://purl.umn.edu/109914</dc:identifier>
<base_dc:link>http://purl.umn.edu/109914</base_dc:link>
<dc:language>en_US</dc:language>
<dc:relation>Project Reports</dc:relation>
<dc:relation>202</dc:relation>
<dc:relation>http://purl.umn.edu/109914</dc:relation>
<base_dc:oa>2</base_dc:oa>
<base_dc:lang>eng</base_dc:lang>
</base_dc:dc>
</metadata>
</record>

View File

@ -0,0 +1,185 @@
<xsl:stylesheet
xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
xmlns:oaire="http://namespace.openaire.eu/schema/oaire/"
xmlns:vocabulary="http://eu/dnetlib/transform/clean"
xmlns:dateCleaner="http://eu/dnetlib/transform/dateISO"
xmlns:oaf="http://namespace.openaire.eu/oaf"
xmlns:oai="http://www.openarchives.org/OAI/2.0/"
xmlns:datacite="http://datacite.org/schema/kernel-4"
xmlns:dri="http://www.driver-repository.eu/namespace/dri"
xmlns:xs="http://www.w3.org/2001/XMLSchema"
xmlns:dr="http://www.driver-repository.eu/namespace/dr"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xmlns:dc="http://purl.org/dc/elements/1.1/"
exclude-result-prefixes="xsl vocabulary dateCleaner" version="2.0">
<xsl:param name="varOfficialName" />
<xsl:param name="varDataSourceId" />
<xsl:param name="varFP7" select="'corda_______::'" />
<xsl:param name="varH2020" select="'corda__h2020::'" />
<xsl:param name="repoCode"
select="substring-before(//*[local-name() = 'header']/*[local-name()='recordIdentifier'], ':')" />
<xsl:param name="index" select="0" />
<xsl:param name="transDate" select="current-dateTime()" />
asdf;klamsdof'sdn &
<xsl:template match="/asdfasdf asdf&">
c:format'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:type" />
<xsl:with-param name="targetElement" select="'dc:type'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:source" />
<xsl:with-param name="targetElement" select="'dc:source'" />
</xsl:call-template>
<dc:language>
<xsl:value-of select="vocabulary:clean( //dc:language, 'dnet:languages')" />
</dc:language>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:rights" />
<xsl:with-param name="targetElement" select="'dc:rights'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:relation[not(starts-with(.,'info:cnr-pdr'))]" />
<xsl:with-param name="targetElement" select="'dc:relation'" />
</xsl:call-template>
<xsl:call-template name="allElements">
<xsl:with-param name="sourceElement" select="//dc:identifier[starts-with(., 'http')]" />
<xsl:with-param name="targetElement" select="'dc:identifier'" />
</xsl:call-template>
<xsl:for-each select="//dc:relation">
<xsl:if test="matches(normalize-space(.), '(info:eu-repo/grantagreement/ec/fp7/)(\d\d\d\d\d\d)(.*)', 'i')">
<oaf:projectid>
<xsl:value-of select="concat($varFP7, replace(normalize-space(.), '(info:eu-repo/grantagreement/ec/fp7/)(\d\d\d\d\d\d)(.*)', '$2', 'i'))" />
</oaf:projectid>
</xsl:if>
<xsl:if test="matches(normalize-space(.), '(info:eu-repo/grantagreement/ec/h2020/)(\d\d\d\d\d\d)(.*)', 'i')">
<oaf:projectid>
<xsl:value-of select="concat($varH2020, replace(normalize-space(.), '(info:eu-repo/grantagreement/ec/h2020/)(\d\d\d\d\d\d)(.*)', '$2', 'i'))" />
</oaf:projectid>
</xsl:if>
</xsl:for-each>
<oaf:accessrights>
<xsl:value-of select="vocabulary:clean( //dc:rights, 'dnet:access_modes')" />
</oaf:accessrights>
<xsl:variable name="varCobjCategory" select="vocabulary:clean( //dc:type, 'dnet:publication_resource')" />
<xsl:variable name="varSuperType" select="vocabulary:clean( $varCobjCategory, 'dnet:result_typologies')" />
<dr:CobjCategory type="{$varSuperType}"><xsl:value-of select="$varCobjCategory" /></dr:CobjCategory>
<xsl:variable name="varRefereedConvt" select="for $i in (//dc:type, //dc:description, //oai:setSpec) return vocabulary:clean( normalize-space($i), 'dnet:review_levels')" />
<xsl:variable name="varRefereedIdntf" select="(//*[string(node-name(.)) = 'dc:identifier' and matches(lower-case(.), '(^|.*[\.\-_/\s\(\)%\d#])pre[\.\-_/\s\(\)%\d#]?prints?([\.\-_/\s\(\)%\d#].*)?$')][count(//dc:identifier) = 1]/'0002', //*[string(node-name(.)) = 'dc:identifier' and matches(lower-case(.), '(^|.*[\.\-_/\s\(\)%\d#])refereed([\.\-_/\s\(\)\d%\d#].*)?$')]/'0001', //*[string(node-name(.)) = 'dc:identifier' and matches(lower-case(.), '.*-peer-reviewed-(fulltext-)?article-.*')]/'0001')" />
<xsl:variable name="varRefereedSourc" select="//*[string(node-name(.)) = ('dc:source', 'dc:publisher') and matches(lower-case(.), '^(.*\s)?pre[\s\-_]*prints?([\s\.,].*)?$')]/'0002'" />
<xsl:variable name="varRefereedDescr" select="(//dc:description[matches(lower-case(.), '.*(this\s*book|this\s*volume|it)\s*constitutes\s*the\s*(thoroughly\s*)?refereed') or matches(lower-case(.), '.*peer[\.\-_/\s\(\)]?review\s*under\s*responsibility\s*of.*') or matches(lower-case(.), '(this|a)\s*(article|preprint)\s*(has\s*been\s*)?(peer[\-\s]*)?reviewed\s*and\s*recommended\s*by\s*peer[\-\s]*community')]/'0001', //dc:description[matches(., '^version\s*(préliminaire.*|preliminary.*|0$)')]/'0002')" />
<xsl:variable name="varRefereedTitle" select="(//dc:title[matches(lower-case(.), '.*\[.*peer[\s\-\._]*review\s*:.*\]\s*$')]/'0001', //dc:title[matches(lower-case(.), '.*\(\s*pre[\s\-\._]*prints?\s*\)\s*$')]/'0002')" />
<xsl:variable name="varRefereedSubjt" select="(//dc:subject[matches(lower-case(.), '^\s*refereed\s*$')][//oaf:datasourceprefix = 'narcis______']/'0001', //dc:subject[matches(lower-case(.), '^\s*no[nt].{0,3}refereed\s*$')][//oaf:datasourceprefix = 'narcis______']/'0002')" />
<xsl:variable name="varRefereed" select="($varRefereedConvt, $varRefereedIdntf, $varRefereedSourc, $varRefereedDescr, $varRefereedTitle, $varRefereedSubjt)" />
<xsl:choose>
<xsl:when test="count($varRefereed[. = '0001']) &gt; 0">
<oaf:refereed>
<xsl:value-of select="'0001'" />
</oaf:refereed>
</xsl:when>
<xsl:when test="count($varRefereed[. = '0002']) &gt; 0">
<oaf:refereed>
<xsl:value-of select="'0002'" />
</oaf:refereed>
</xsl:when>
</xsl:choose>
<oaf:dateAccepted>
<xsl:value-of select="dateCleaner:dateISO( //dc:date[1] )" />
</oaf:dateAccepted>
<xsl:if test="//dc:relation[starts-with(., 'http')] and //dc:rights[.='info:eu-repo/semantics/openAccess']">
<oaf:fulltext>
<xsl:value-of select="//dc:relation[starts-with(., 'http')]" />
</oaf:fulltext>
</xsl:if>
<oaf:hostedBy name="{$varOfficialName}" id="{$varDataSourceId}" />
<oaf:collectedFrom name="{$varOfficialName}" id="{$varDataSourceId}" />
<xsl:variable name="varKnownFileEndings" select="('.bmp', '.doc', '.docx', '.epub', '.flv', '.jpeg', '.jpg', '.m4v', '.mp4', '.mpg', '.odp', '.pdf', '.png', '.ppt', '.tiv', '.txt', '.xls', '.xlsx', '.zip')" />
<xsl:variable name="varIdDoi" select="distinct-values((//dc:identifier[starts-with(., '10.')][matches(., '(10[.][0-9]{4,}[^\s/&gt;]*/[^\s&gt;]+)')], //dc:identifier[starts-with(., 'http') and (contains(., '://dx.doi.org/10.') or contains(., '://doi.org/10.'))]/substring-after(., 'doi.org/'), //dc:identifier[starts-with(lower-case(.), 'doi:10.')]/substring-after(lower-case(.), 'doi:')))" />
<xsl:for-each select="$varIdDoi">
<oaf:identifier identifierType="doi">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:variable name="varIdHdl" select="distinct-values(//dc:identifier[starts-with(., 'http') and contains(., '://hdl.handle.net/')]/substring-after(., 'hdl.handle.net/'))" />
<xsl:for-each select="$varIdHdl" >
<oaf:identifier identifierType="handle">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:variable name="varIdUrn" select="distinct-values(//dc:identifier[starts-with(., 'urn:nbn:nl:') or starts-with(., 'URN:NBN:NL:')])" />
<xsl:for-each select="$varIdUrn">
<oaf:identifier identifierType="urn">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:variable name="varOrigBaseUrl" select="//*[local-name() = 'about']/*[local-name() = 'provenance']//*[local-name() = 'originDescription' and not(./*[local-name() = 'originDescription'])]/*[local-name() = 'baseURL']" />
<xsl:variable name="varIdLdpg" select="distinct-values(//dc:identifier[(contains(substring-after(., '://'), '/') and contains($varOrigBaseUrl, substring-before(substring-after(., '://'), '/'))) or (contains(substring-after(., '://'), ':') and contains($varOrigBaseUrl, substring-before(substring-after(., '://'), ':')))][not(replace(lower-case(.), '.*(\.[a-z]*)$', '$1') = $varKnownFileEndings)])" />
<xsl:for-each select="$varIdLdpg">
<oaf:identifier identifierType="landingPage">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:variable name="varIdUrl" select="distinct-values(//dc:identifier[starts-with(., 'http')][not(contains(., '://dx.doi.org/') or contains(., '://doi.org/') or contains(., '://hdl.handle.net/'))][count(index-of($varIdLdpg, .)) = 0])" />
<xsl:for-each select="$varIdUrl">
<oaf:identifier identifierType="url">
<xsl:value-of select="." />
</oaf:identifier>
</xsl:for-each>
<xsl:for-each select="//oai:setSpec">
<xsl:variable name="rorDsId" select="vocabulary:clean(., 'cnr:institutes')" />
<xsl:if test="contains($rorDsId, '/ror.org/')">
<oaf:relation relType="resultOrganization" subRelType="affiliation" relClass="hasAuthorInstitution">
<xsl:value-of select="concat('ror_________::', $rorDsId)" />
</oaf:relation>
</xsl:if>
</xsl:for-each>
</metadata>
<xsl:copy-of select="//*[local-name() = 'about']" />
</record>
</xsl:template>
<xsl:template name="allElements">
<xsl:param name="sourceElement" />
<xsl:param name="targetElement" />
<xsl:for-each select="$sourceElement">
<xsl:element name="{$targetElement}">
<xsl:value-of select="normalize-space(.)" />
</xsl:element>
</xsl:for-each>
</xsl:template>
<xsl:template match="//*[local-name() = 'header']">
<xsl:copy>
<xsl:apply-templates select="node()|@*" />
<xsl:element name="dr:dateOfTransformation">
<xsl:value-of select="$transDate" />
</xsl:element>
</xsl:copy>
</xsl:template>
<xsl:template match="node()|@*">
<xsl:copy>
<xsl:apply-templates select="node()|@*" />
</xsl:copy>
</xsl:template>
</xsl:stylesheet>

View File

@ -1496,4 +1496,30 @@ cnr:institutes @=@ __CDS131__ @=@ IBE - Istituto per la BioEconomia
cnr:institutes @=@ https://ror.org/0263zy895 @=@ CDS132
cnr:institutes @=@ https://ror.org/0263zy895 @=@ SCITEC - Istituto di Scienze e Tecnologie Chimiche \"Giulio Natta\"
cnr:institutes @=@ __CDS133__ @=@ CDS133
cnr:institutes @=@ __CDS133__ @=@ STEMS - Istituto di Scienze e Tecnologie per l'Energia e la Mobilità Sostenibili
cnr:institutes @=@ __CDS133__ @=@ STEMS - Istituto di Scienze e Tecnologie per l'Energia e la Mobilità Sostenibili
base:normalized_types @=@ Text @=@ 1
base:normalized_types @=@ Book @=@ 11
base:normalized_types @=@ Book part @=@ 111
base:normalized_types @=@ Journal/Newspaper @=@ 12
base:normalized_types @=@ Article contribution @=@ 121
base:normalized_types @=@ Other non-article @=@ 122
base:normalized_types @=@ Conference object @=@ 13
base:normalized_types @=@ Report @=@ 14
base:normalized_types @=@ Review @=@ 15
base:normalized_types @=@ Course material @=@ 16
base:normalized_types @=@ Lecture @=@ 17
base:normalized_types @=@ Thesis @=@ 18
base:normalized_types @=@ Bachelor's thesis @=@ 181
base:normalized_types @=@ Master's thesis @=@ 182
base:normalized_types @=@ Doctoral and postdoctoral thesis @=@ 183
base:normalized_types @=@ Manuscript @=@ 19
base:normalized_types @=@ Patent @=@ 1A
base:normalized_types @=@ Musical notation @=@ 2
base:normalized_types @=@ Map @=@ 3
base:normalized_types @=@ Audio @=@ 4
base:normalized_types @=@ Image/Video @=@ 5
base:normalized_types @=@ Still image @=@ 51
base:normalized_types @=@ Moving image/Video @=@ 52
base:normalized_types @=@ Software @=@ 6
base:normalized_types @=@ Dataset @=@ 7
base:normalized_types @=@ Unknown @=@ F

View File

@ -1210,4 +1210,29 @@ cnr:institutes @=@ cnr:institutes @=@ __CDS130__ @=@ __CDS130__
cnr:institutes @=@ cnr:institutes @=@ __CDS131__ @=@ __CDS131__
cnr:institutes @=@ cnr:institutes @=@ https://ror.org/0263zy895 @=@ https://ror.org/0263zy895
cnr:institutes @=@ cnr:institutes @=@ __CDS133__ @=@ __CDS133__
base:normalized_types @=@ base:normalized_types @=@ Text @=@ Text
base:normalized_types @=@ base:normalized_types @=@ Book @=@ Book
base:normalized_types @=@ base:normalized_types @=@ Book part @=@ Book part
base:normalized_types @=@ base:normalized_types @=@ Journal/Newspaper @=@ Journal/Newspaper
base:normalized_types @=@ base:normalized_types @=@ Article contribution @=@ Article contribution
base:normalized_types @=@ base:normalized_types @=@ Other non-article @=@ Other non-article
base:normalized_types @=@ base:normalized_types @=@ Conference object @=@ Conference object
base:normalized_types @=@ base:normalized_types @=@ Report @=@ Report
base:normalized_types @=@ base:normalized_types @=@ Review @=@ Review
base:normalized_types @=@ base:normalized_types @=@ Course material @=@ Course material
base:normalized_types @=@ base:normalized_types @=@ Lecture @=@ Lecture
base:normalized_types @=@ base:normalized_types @=@ Thesis @=@ Thesis
base:normalized_types @=@ base:normalized_types @=@ Bachelor's thesis @=@ Bachelor's thesis
base:normalized_types @=@ base:normalized_types @=@ Master's thesis @=@ Master's thesis
base:normalized_types @=@ base:normalized_types @=@ Doctoral and postdoctoral thesis @=@ Doctoral and postdoctoral thesis
base:normalized_types @=@ base:normalized_types @=@ Manuscript @=@ Manuscript
base:normalized_types @=@ base:normalized_types @=@ Patent @=@ Patent
base:normalized_types @=@ base:normalized_types @=@ Musical notation @=@ Musical notation
base:normalized_types @=@ base:normalized_types @=@ Map @=@ Map
base:normalized_types @=@ base:normalized_types @=@ Audio @=@ Audio
base:normalized_types @=@ base:normalized_types @=@ Image/Video @=@ Image/Video
base:normalized_types @=@ base:normalized_types @=@ Still image @=@ Still image
base:normalized_types @=@ base:normalized_types @=@ Moving image/Video @=@ Moving image/Video
base:normalized_types @=@ base:normalized_types @=@ Software @=@ Software
base:normalized_types @=@ base:normalized_types @=@ Dataset @=@ Dataset
base:normalized_types @=@ base:normalized_types @=@ Unknown @=@ Unknown

View File

@ -4,6 +4,7 @@ package eu.dnetlib.dhp.oa.dedup;
import java.util.*;
import java.util.stream.Stream;
import eu.dnetlib.dhp.schema.oaf.utils.MergeUtils;
import org.apache.commons.beanutils.BeanUtils;
import org.apache.commons.lang3.StringUtils;
import org.apache.spark.api.java.function.FlatMapFunction;
@ -14,6 +15,7 @@ import org.apache.spark.sql.*;
import eu.dnetlib.dhp.oa.dedup.model.Identifier;
import eu.dnetlib.dhp.oa.merge.AuthorMerger;
import eu.dnetlib.dhp.schema.common.ModelSupport;
import eu.dnetlib.dhp.schema.oaf.utils.MergeUtils;
import eu.dnetlib.dhp.schema.oaf.Author;
import eu.dnetlib.dhp.schema.oaf.DataInfo;
import eu.dnetlib.dhp.schema.oaf.OafEntity;
@ -23,190 +25,190 @@ import scala.Tuple3;
import scala.collection.JavaConversions;
public class DedupRecordFactory {
public static final class DedupRecordReduceState {
public final String dedupId;
public static final class DedupRecordReduceState {
public final String dedupId;
public final ArrayList<String> aliases = new ArrayList<>();
public final ArrayList<String> aliases = new ArrayList<>();
public final HashSet<String> acceptanceDate = new HashSet<>();
public final HashSet<String> acceptanceDate = new HashSet<>();
public OafEntity entity;
public OafEntity entity;
public DedupRecordReduceState(String dedupId, String id, OafEntity entity) {
this.dedupId = dedupId;
this.entity = entity;
if (entity == null) {
aliases.add(id);
} else {
if (Result.class.isAssignableFrom(entity.getClass())) {
Result result = (Result) entity;
if (result.getDateofacceptance() != null
&& StringUtils.isNotBlank(result.getDateofacceptance().getValue())) {
acceptanceDate.add(result.getDateofacceptance().getValue());
}
}
}
}
public DedupRecordReduceState(String dedupId, String id, OafEntity entity) {
this.dedupId = dedupId;
this.entity = entity;
if (entity == null) {
aliases.add(id);
} else {
if (Result.class.isAssignableFrom(entity.getClass())) {
Result result = (Result) entity;
if (result.getDateofacceptance() != null
&& StringUtils.isNotBlank(result.getDateofacceptance().getValue())) {
acceptanceDate.add(result.getDateofacceptance().getValue());
}
}
}
}
public String getDedupId() {
return dedupId;
}
}
public String getDedupId() {
return dedupId;
}
}
private static final int MAX_ACCEPTANCE_DATE = 20;
private static final int MAX_ACCEPTANCE_DATE = 20;
private DedupRecordFactory() {
}
private DedupRecordFactory() {
}
public static Dataset<OafEntity> createDedupRecord(
final SparkSession spark,
final DataInfo dataInfo,
final String mergeRelsInputPath,
final String entitiesInputPath,
final Class<OafEntity> clazz) {
public static Dataset<OafEntity> createDedupRecord(
final SparkSession spark,
final DataInfo dataInfo,
final String mergeRelsInputPath,
final String entitiesInputPath,
final Class<OafEntity> clazz) {
final long ts = System.currentTimeMillis();
final Encoder<OafEntity> beanEncoder = Encoders.bean(clazz);
final Encoder<OafEntity> kryoEncoder = Encoders.kryo(clazz);
final long ts = System.currentTimeMillis();
final Encoder<OafEntity> beanEncoder = Encoders.bean(clazz);
final Encoder<OafEntity> kryoEncoder = Encoders.kryo(clazz);
// <id, json_entity>
Dataset<Row> entities = spark
.read()
.schema(Encoders.bean(clazz).schema())
.json(entitiesInputPath)
.as(beanEncoder)
.map(
(MapFunction<OafEntity, Tuple2<String, OafEntity>>) entity -> {
return new Tuple2<>(entity.getId(), entity);
},
Encoders.tuple(Encoders.STRING(), kryoEncoder))
.selectExpr("_1 AS id", "_2 AS kryoObject");
// <id, json_entity>
Dataset<Row> entities = spark
.read()
.schema(Encoders.bean(clazz).schema())
.json(entitiesInputPath)
.as(beanEncoder)
.map(
(MapFunction<OafEntity, Tuple2<String, OafEntity>>) entity -> {
return new Tuple2<>(entity.getId(), entity);
},
Encoders.tuple(Encoders.STRING(), kryoEncoder))
.selectExpr("_1 AS id", "_2 AS kryoObject");
// <source, target>: source is the dedup_id, target is the id of the mergedIn
Dataset<Row> mergeRels = spark
.read()
.load(mergeRelsInputPath)
.where("relClass == 'merges'")
.selectExpr("source as dedupId", "target as id");
// <source, target>: source is the dedup_id, target is the id of the mergedIn
Dataset<Row> mergeRels = spark
.read()
.load(mergeRelsInputPath)
.where("relClass == 'merges'")
.selectExpr("source as dedupId", "target as id");
return mergeRels
.join(entities, JavaConversions.asScalaBuffer(Collections.singletonList("id")), "left")
.select("dedupId", "id", "kryoObject")
.as(Encoders.tuple(Encoders.STRING(), Encoders.STRING(), kryoEncoder))
.map(
(MapFunction<Tuple3<String, String, OafEntity>, DedupRecordReduceState>) t -> new DedupRecordReduceState(
t._1(), t._2(), t._3()),
Encoders.kryo(DedupRecordReduceState.class))
.groupByKey(
(MapFunction<DedupRecordReduceState, String>) DedupRecordReduceState::getDedupId, Encoders.STRING())
.reduceGroups(
(ReduceFunction<DedupRecordReduceState>) (t1, t2) -> {
if (t1.entity == null) {
t2.aliases.addAll(t1.aliases);
return t2;
}
if (t1.acceptanceDate.size() < MAX_ACCEPTANCE_DATE) {
t1.acceptanceDate.addAll(t2.acceptanceDate);
}
t1.aliases.addAll(t2.aliases);
t1.entity = reduceEntity(t1.entity, t2.entity);
return mergeRels
.join(entities, JavaConversions.asScalaBuffer(Collections.singletonList("id")), "left")
.select("dedupId", "id", "kryoObject")
.as(Encoders.tuple(Encoders.STRING(), Encoders.STRING(), kryoEncoder))
.map(
(MapFunction<Tuple3<String, String, OafEntity>, DedupRecordReduceState>) t -> new DedupRecordReduceState(
t._1(), t._2(), t._3()),
Encoders.kryo(DedupRecordReduceState.class))
.groupByKey(
(MapFunction<DedupRecordReduceState, String>) DedupRecordReduceState::getDedupId, Encoders.STRING())
.reduceGroups(
(ReduceFunction<DedupRecordReduceState>) (t1, t2) -> {
if (t1.entity == null) {
t2.aliases.addAll(t1.aliases);
return t2;
}
if (t1.acceptanceDate.size() < MAX_ACCEPTANCE_DATE) {
t1.acceptanceDate.addAll(t2.acceptanceDate);
}
t1.aliases.addAll(t2.aliases);
t1.entity = reduceEntity(t1.entity, t2.entity);
return t1;
})
.flatMap((FlatMapFunction<Tuple2<String, DedupRecordReduceState>, OafEntity>) t -> {
String dedupId = t._1();
DedupRecordReduceState agg = t._2();
return t1;
})
.flatMap((FlatMapFunction<Tuple2<String, DedupRecordReduceState>, OafEntity>) t -> {
String dedupId = t._1();
DedupRecordReduceState agg = t._2();
if (agg.acceptanceDate.size() >= MAX_ACCEPTANCE_DATE) {
return Collections.emptyIterator();
}
if (agg.acceptanceDate.size() >= MAX_ACCEPTANCE_DATE) {
return Collections.emptyIterator();
}
return Stream
.concat(
Stream
.of(agg.getDedupId())
.map(id -> createDedupOafEntity(id, agg.entity, dataInfo, ts)),
agg.aliases
.stream()
.map(id -> createMergedDedupAliasOafEntity(id, agg.entity, dataInfo, ts)))
.iterator();
}, beanEncoder);
}
return Stream
.concat(
Stream
.of(agg.getDedupId())
.map(id -> createDedupOafEntity(id, agg.entity, dataInfo, ts)),
agg.aliases
.stream()
.map(id -> createMergedDedupAliasOafEntity(id, agg.entity, dataInfo, ts)))
.iterator();
}, beanEncoder);
}
private static OafEntity createDedupOafEntity(String id, OafEntity base, DataInfo dataInfo, long ts) {
try {
OafEntity res = (OafEntity) BeanUtils.cloneBean(base);
res.setId(id);
res.setDataInfo(dataInfo);
res.setLastupdatetimestamp(ts);
return res;
} catch (Exception e) {
throw new RuntimeException(e);
}
}
private static OafEntity createDedupOafEntity(String id, OafEntity base, DataInfo dataInfo, long ts) {
try {
OafEntity res = (OafEntity) BeanUtils.cloneBean(base);
res.setId(id);
res.setDataInfo(dataInfo);
res.setLastupdatetimestamp(ts);
return res;
} catch (Exception e) {
throw new RuntimeException(e);
}
}
private static OafEntity createMergedDedupAliasOafEntity(String id, OafEntity base, DataInfo dataInfo, long ts) {
try {
OafEntity res = createDedupOafEntity(id, base, dataInfo, ts);
DataInfo ds = (DataInfo) BeanUtils.cloneBean(dataInfo);
ds.setDeletedbyinference(true);
res.setDataInfo(ds);
return res;
} catch (Exception e) {
throw new RuntimeException(e);
}
}
private static OafEntity createMergedDedupAliasOafEntity(String id, OafEntity base, DataInfo dataInfo, long ts) {
try {
OafEntity res = createDedupOafEntity(id, base, dataInfo, ts);
DataInfo ds = (DataInfo) BeanUtils.cloneBean(dataInfo);
ds.setDeletedbyinference(true);
res.setDataInfo(ds);
return res;
} catch (Exception e) {
throw new RuntimeException(e);
}
}
private static OafEntity reduceEntity(OafEntity entity, OafEntity duplicate) {
private static OafEntity reduceEntity(OafEntity entity, OafEntity duplicate) {
if (duplicate == null) {
return entity;
}
if (duplicate == null) {
return entity;
}
int compare = new IdentifierComparator<>()
.compare(Identifier.newInstance(entity), Identifier.newInstance(duplicate));
int compare = new IdentifierComparator<>()
.compare(Identifier.newInstance(entity), Identifier.newInstance(duplicate));
if (compare > 0) {
OafEntity swap = duplicate;
duplicate = entity;
entity = swap;
}
if (compare > 0) {
OafEntity swap = duplicate;
duplicate = entity;
entity = swap;
}
entity.mergeFrom(duplicate);
entity = MergeUtils.checkedMerge(entity, duplicate);
if (ModelSupport.isSubClass(duplicate, Result.class)) {
Result re = (Result) entity;
Result rd = (Result) duplicate;
if (ModelSupport.isSubClass(duplicate, Result.class)) {
Result re = (Result) entity;
Result rd = (Result) duplicate;
List<List<Author>> authors = new ArrayList<>();
if (re.getAuthor() != null) {
authors.add(re.getAuthor());
}
if (rd.getAuthor() != null) {
authors.add(rd.getAuthor());
}
List<List<Author>> authors = new ArrayList<>();
if (re.getAuthor() != null) {
authors.add(re.getAuthor());
}
if (rd.getAuthor() != null) {
authors.add(rd.getAuthor());
}
re.setAuthor(AuthorMerger.merge(authors));
}
re.setAuthor(AuthorMerger.merge(authors));
}
return entity;
}
return entity;
}
public static <T extends OafEntity> T entityMerger(
String id, Iterator<Tuple2<String, T>> entities, long ts, DataInfo dataInfo, Class<T> clazz) {
T base = entities.next()._2();
public static <T extends OafEntity> T entityMerger(
String id, Iterator<Tuple2<String, T>> entities, long ts, DataInfo dataInfo, Class<T> clazz) {
T base = entities.next()._2();
while (entities.hasNext()) {
T duplicate = entities.next()._2();
if (duplicate != null)
base = (T) reduceEntity(base, duplicate);
}
while (entities.hasNext()) {
T duplicate = entities.next()._2();
if (duplicate != null)
base = (T) reduceEntity(base, duplicate);
}
base.setId(id);
base.setDataInfo(dataInfo);
base.setLastupdatetimestamp(ts);
base.setId(id);
base.setDataInfo(dataInfo);
base.setLastupdatetimestamp(ts);
return base;
}
return base;
}
}

View File

@ -3,6 +3,7 @@ package eu.dnetlib.dhp.oa.dedup;
import static org.apache.spark.sql.functions.col;
import eu.dnetlib.dhp.schema.oaf.utils.MergeUtils;
import org.apache.commons.io.IOUtils;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.function.MapFunction;
@ -127,10 +128,8 @@ public class SparkPropagateRelation extends AbstractSparkAction {
(MapFunction<Relation, String>) r -> String
.join(" ", r.getSource(), r.getTarget(), r.getRelType(), r.getSubRelType(), r.getRelClass()),
Encoders.STRING())
.reduceGroups((ReduceFunction<Relation>) (b, a) -> {
b.mergeFrom(a);
return b;
})
.reduceGroups((ReduceFunction<Relation>) MergeUtils::mergeRelation
)
.map((MapFunction<Tuple2<String, Relation>, Relation>) Tuple2::_2, REL_BEAN_ENC);
final String outputRelationPath = graphOutputPath + "/relation";

View File

@ -3,17 +3,18 @@ package eu.dnetlib.doiboost
import eu.dnetlib.dhp.application.ArgumentApplicationParser
import eu.dnetlib.dhp.oa.merge.AuthorMerger
import eu.dnetlib.dhp.schema.common.ModelConstants
import eu.dnetlib.dhp.schema.oaf.utils.MergeUtils
import eu.dnetlib.dhp.schema.oaf.{Organization, Publication, Relation, Dataset => OafDataset}
import eu.dnetlib.doiboost.mag.ConversionUtil
import org.apache.commons.io.IOUtils
import org.apache.spark.SparkConf
import org.apache.spark.sql.expressions.Aggregator
import org.apache.spark.sql.functions.col
import org.apache.spark.sql._
import org.json4s.DefaultFormats
import org.json4s.JsonAST.{JField, JObject, JString}
import org.json4s.jackson.JsonMethods.parse
import org.slf4j.{Logger, LoggerFactory}
import scala.collection.JavaConverters._
object SparkGenerateDoiBoost {
@ -78,8 +79,10 @@ object SparkGenerateDoiBoost {
if (item._2 != null) {
val otherPub = item._2._2
if (otherPub != null) {
crossrefPub.mergeFrom(otherPub)
crossrefPub.setAuthor(AuthorMerger.mergeAuthor(crossrefPub.getAuthor, otherPub.getAuthor))
val mergedAuthor = AuthorMerger.mergeAuthor(crossrefPub.getAuthor, otherPub.getAuthor)
val res = MergeUtils.mergePublication(crossrefPub, otherPub)
res.setAuthor(mergedAuthor);
return res
}
}
crossrefPub
@ -130,14 +133,13 @@ object SparkGenerateDoiBoost {
// So we have to merge
val b1 = left._2
val b2 = right._2
b1.mergeFrom(b2)
b1.mergeOAFDataInfo(b2)
val authors = AuthorMerger.mergeAuthor(b1.getAuthor, b2.getAuthor)
b1.setAuthor(authors)
val merged = MergeUtils.mergePublication(b1, b2);
merged.setAuthor(authors)
if (b2.getId != null && b2.getId.nonEmpty)
b1.setId(b2.getId)
merged.setId(b2.getId)
//Return publication Merged
(b1.getId, b1)
(merged.getId, merged)
}
} else {
// Left is Null so we return right

View File

@ -1,8 +1,8 @@
package eu.dnetlib.doiboost.mag
import eu.dnetlib.dhp.schema.common.ModelConstants
import eu.dnetlib.dhp.schema.oaf.utils.IdentifierFactory
import eu.dnetlib.dhp.schema.oaf.{Instance, Journal, Publication, StructuredProperty, Subject}
import eu.dnetlib.dhp.schema.oaf.utils.{IdentifierFactory, MergeUtils}
import eu.dnetlib.dhp.schema.oaf.{Instance, Journal, Publication, Subject}
import eu.dnetlib.doiboost.DoiBoostMappingUtil
import eu.dnetlib.doiboost.DoiBoostMappingUtil._
import org.json4s
@ -142,8 +142,7 @@ case object ConversionUtil {
def mergePublication(a: Publication, b: Publication): Publication = {
if ((a != null) && (b != null)) {
a.mergeFrom(b)
a
MergeUtils.mergePublication(a, b)
} else {
if (a == null) b else a
}

View File

@ -0,0 +1,39 @@
package eu.dnetlib.dhp.bulktag.actions;
import java.io.Serializable;
import java.util.List;
/**
* @author miriam.baglioni
* @Date 22/01/24
*/
public class Action implements Serializable {
private String clazz;
private String method;
private List<Parameters> params;
public String getClazz() {
return clazz;
}
public void setClazz(String clazz) {
this.clazz = clazz;
}
public String getMethod() {
return method;
}
public void setMethod(String method) {
this.method = method;
}
public List<Parameters> getParams() {
return params;
}
public void setParams(List<Parameters> params) {
this.params = params;
}
}

View File

@ -0,0 +1,45 @@
package eu.dnetlib.dhp.bulktag.actions;
import java.io.Serializable;
/**
* @author miriam.baglioni
* @Date 19/01/24
*/
public class ExecSubstringAction implements Serializable {
private String value;
private String from;
private String to;
public String getValue() {
return value;
}
public void setValue(String value) {
this.value = value;
}
public String getFrom() {
return from;
}
public void setFrom(String from) {
this.from = from;
}
public String getTo() {
return to;
}
public void setTo(String to) {
this.to = to;
}
public String execSubstring() {
return this.value.substring(Integer.valueOf(this.from), Integer.valueOf(this.to));
}
}

View File

@ -0,0 +1,30 @@
package eu.dnetlib.dhp.bulktag.actions;
import java.io.Serializable;
/**
* @author miriam.baglioni
* @Date 22/01/24
*/
public class MapModel implements Serializable {
private String path;
private Action action;
public String getPath() {
return path;
}
public void setPath(String path) {
this.path = path;
}
public Action getAction() {
return action;
}
public void setAction(Action action) {
this.action = action;
}
}

View File

@ -0,0 +1,29 @@
package eu.dnetlib.dhp.bulktag.actions;
import java.io.Serializable;
/**
* @author miriam.baglioni
* @Date 22/01/24
*/
public class Parameters implements Serializable {
private String paramName;
private String paramValue;
public String getParamName() {
return paramName;
}
public void setParamName(String paramName) {
this.paramName = paramName;
}
public String getParamValue() {
return paramValue;
}
public void setParamValue(String paramValue) {
this.paramValue = paramValue;
}
}

View File

@ -4,7 +4,9 @@ package eu.dnetlib.dhp.bulktag.community;
import java.io.Serializable;
import java.util.HashMap;
public class ProtoMap extends HashMap<String, String> implements Serializable {
import eu.dnetlib.dhp.bulktag.actions.MapModel;
public class ProtoMap extends HashMap<String, MapModel> implements Serializable {
public ProtoMap() {
super();

View File

@ -5,6 +5,8 @@ import static eu.dnetlib.dhp.bulktag.community.TaggingConstants.*;
import static eu.dnetlib.dhp.schema.common.ModelConstants.*;
import java.io.Serializable;
import java.lang.reflect.InvocationTargetException;
import java.lang.reflect.Method;
import java.util.*;
import java.util.stream.Collectors;
@ -15,7 +17,10 @@ import org.slf4j.LoggerFactory;
import com.google.gson.Gson;
import com.jayway.jsonpath.DocumentContext;
import com.jayway.jsonpath.JsonPath;
import com.jayway.jsonpath.PathNotFoundException;
import eu.dnetlib.dhp.bulktag.actions.MapModel;
import eu.dnetlib.dhp.bulktag.actions.Parameters;
import eu.dnetlib.dhp.bulktag.eosc.EoscIFTag;
import eu.dnetlib.dhp.schema.oaf.*;
import eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils;
@ -35,27 +40,59 @@ public class ResultTagger implements Serializable {
return (tmp != clist.size());
}
private Map<String, List<String>> getParamMap(final Result result, Map<String, String> params) {
private Map<String, List<String>> getParamMap(final Result result, Map<String, MapModel> params)
throws NoSuchMethodException, InvocationTargetException {
Map<String, List<String>> param = new HashMap<>();
String json = new Gson().toJson(result, Result.class);
DocumentContext jsonContext = JsonPath.parse(json);
if (params == null) {
params = new HashMap<>();
}
for (String key : params.keySet()) {
MapModel mapModel = params.get(key);
try {
param.put(key, jsonContext.read(params.get(key)));
} catch (com.jayway.jsonpath.PathNotFoundException e) {
String path = mapModel.getPath();
Object obj = jsonContext.read(path);
List<String> pathValue;
if (obj instanceof java.lang.String)
pathValue = Arrays.asList((String) obj);
else
pathValue = (List<String>) obj;
if (Optional.ofNullable(mapModel.getAction()).isPresent()) {
Class<?> c = Class.forName(mapModel.getAction().getClazz());
Object class_instance = c.newInstance();
Method setField = c.getMethod("setValue", String.class);
setField.invoke(class_instance, pathValue.get(0));
for (Parameters p : mapModel.getAction().getParams()) {
setField = c.getMethod("set" + p.getParamName(), String.class);
setField.invoke(class_instance, p.getParamValue());
}
param
.put(
key, Arrays
.asList((String) c.getMethod(mapModel.getAction().getMethod()).invoke(class_instance)));
}
else {
param.put(key, pathValue);
}
} catch (PathNotFoundException | ClassNotFoundException | InstantiationException
| IllegalAccessException e) {
param.put(key, new ArrayList<>());
}
}
return param;
}
public <R extends Result> R enrichContextCriteria(
final R result, final CommunityConfiguration conf, final Map<String, String> criteria) {
final Map<String, List<String>> param = getParamMap(result, criteria);
final R result, final CommunityConfiguration conf, final Map<String, MapModel> criteria)
throws InvocationTargetException, NoSuchMethodException {
// Verify if the entity is deletedbyinference. In case verify if to clean the context list
// from all the zenodo communities
@ -64,6 +101,8 @@ public class ResultTagger implements Serializable {
return result;
}
final Map<String, List<String>> param = getParamMap(result, criteria);
// Execute the EOSCTag for the services
switch (result.getResulttype().getClassid()) {
case PUBLICATION_RESULTTYPE_CLASSID:

View File

@ -0,0 +1,34 @@
package eu.dnetlib.dhp.bulktag.criteria;
import java.io.Serializable;
/**
* @author miriam.baglioni
* @Date 11/11/22
*/
@VerbClass("greater_than")
public class GreatThanVerb implements Selection, Serializable {
private String param;
public GreatThanVerb() {
}
public GreatThanVerb(final String param) {
this.param = param;
}
@Override
public boolean apply(String value) {
return value.compareTo(param) > 0;
}
public String getParam() {
return param;
}
public void setParam(String param) {
this.param = param;
}
}

View File

@ -0,0 +1,34 @@
package eu.dnetlib.dhp.bulktag.criteria;
import java.io.Serializable;
/**
* @author miriam.baglioni
* @Date 11/11/22
*/
@VerbClass("lesser_than")
public class LessThanVerb implements Selection, Serializable {
private String param;
public LessThanVerb() {
}
public LessThanVerb(final String param) {
this.param = param;
}
@Override
public boolean apply(String value) {
return value.compareTo(param) < 0;
}
public String getParam() {
return param;
}
public void setParam(String param) {
this.param = param;
}
}

View File

@ -149,7 +149,7 @@ public class SparkResultToCommunityFromOrganizationJob {
}
}
// res.setContext(propagatedContexts);
// ret.mergeFrom(res);
// return MergeUtils.mergeResult(ret, res);
}
return ret;
};

View File

@ -13,6 +13,7 @@ import java.util.List;
import java.util.Optional;
import java.util.stream.Collectors;
import eu.dnetlib.dhp.schema.oaf.utils.MergeUtils;
import org.apache.commons.io.IOUtils;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.function.MapFunction;
@ -24,8 +25,6 @@ import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.resulttocommunityfromorganization.ResultCommunityList;
import eu.dnetlib.dhp.resulttocommunityfromorganization.SparkResultToCommunityFromOrganizationJob;
import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.common.ModelSupport;
import eu.dnetlib.dhp.schema.oaf.Context;
@ -162,7 +161,7 @@ public class SparkResultToCommunityFromProject implements Serializable {
}
}
res.setContext(propagatedContexts);
ret.mergeFrom(res);
return MergeUtils.checkedMerge(ret, res);
}
return ret;
};

View File

@ -25,118 +25,118 @@ import scala.Tuple2;
public class SparkResultToCommunityThroughSemRelJob {
private static final Logger log = LoggerFactory.getLogger(SparkResultToCommunityThroughSemRelJob.class);
private static final Logger log = LoggerFactory.getLogger(SparkResultToCommunityThroughSemRelJob.class);
public static void main(String[] args) throws Exception {
public static void main(String[] args) throws Exception {
String jsonConfiguration = IOUtils
.toString(
SparkResultToCommunityThroughSemRelJob.class
.getResourceAsStream(
"/eu/dnetlib/dhp/wf/subworkflows/resulttocommunityfromsemrel/input_communitytoresult_parameters.json"));
String jsonConfiguration = IOUtils
.toString(
SparkResultToCommunityThroughSemRelJob.class
.getResourceAsStream(
"/eu/dnetlib/dhp/wf/subworkflows/resulttocommunityfromsemrel/input_communitytoresult_parameters.json"));
final ArgumentApplicationParser parser = new ArgumentApplicationParser(jsonConfiguration);
final ArgumentApplicationParser parser = new ArgumentApplicationParser(jsonConfiguration);
parser.parseArgument(args);
parser.parseArgument(args);
Boolean isSparkSessionManaged = isSparkSessionManaged(parser);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
Boolean isSparkSessionManaged = isSparkSessionManaged(parser);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
String inputPath = parser.get("sourcePath");
log.info("inputPath: {}", inputPath);
String inputPath = parser.get("sourcePath");
log.info("inputPath: {}", inputPath);
final String outputPath = parser.get("outputPath");
log.info("outputPath: {}", outputPath);
final String outputPath = parser.get("outputPath");
log.info("outputPath: {}", outputPath);
final String preparedInfoPath = parser.get("preparedInfoPath");
log.info("preparedInfoPath: {}", preparedInfoPath);
final String preparedInfoPath = parser.get("preparedInfoPath");
log.info("preparedInfoPath: {}", preparedInfoPath);
SparkConf conf = new SparkConf();
conf.set("hive.metastore.uris", parser.get("hive_metastore_uris"));
SparkConf conf = new SparkConf();
conf.set("hive.metastore.uris", parser.get("hive_metastore_uris"));
final String resultClassName = parser.get("resultTableName");
log.info("resultTableName: {}", resultClassName);
final String resultClassName = parser.get("resultTableName");
log.info("resultTableName: {}", resultClassName);
final Boolean saveGraph = Optional
.ofNullable(parser.get("saveGraph"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("saveGraph: {}", saveGraph);
final Boolean saveGraph = Optional
.ofNullable(parser.get("saveGraph"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("saveGraph: {}", saveGraph);
@SuppressWarnings("unchecked")
Class<? extends Result> resultClazz = (Class<? extends Result>) Class.forName(resultClassName);
@SuppressWarnings("unchecked")
Class<? extends Result> resultClazz = (Class<? extends Result>) Class.forName(resultClassName);
runWithSparkHiveSession(
conf,
isSparkSessionManaged,
spark -> {
if (isTest(parser)) {
removeOutputDir(spark, outputPath);
}
if (saveGraph) {
execPropagation(
spark, inputPath, outputPath, preparedInfoPath, resultClazz);
}
});
}
runWithSparkHiveSession(
conf,
isSparkSessionManaged,
spark -> {
if (isTest(parser)) {
removeOutputDir(spark, outputPath);
}
if (saveGraph) {
execPropagation(
spark, inputPath, outputPath, preparedInfoPath, resultClazz);
}
});
}
private static <R extends Result> void execPropagation(
SparkSession spark,
String inputPath,
String outputPath,
String preparedInfoPath,
Class<R> resultClazz) {
private static <R extends Result> void execPropagation(
SparkSession spark,
String inputPath,
String outputPath,
String preparedInfoPath,
Class<R> resultClazz) {
Dataset<ResultCommunityList> possibleUpdates = readPath(spark, preparedInfoPath, ResultCommunityList.class);
Dataset<R> result = readPath(spark, inputPath, resultClazz);
Dataset<ResultCommunityList> possibleUpdates = readPath(spark, preparedInfoPath, ResultCommunityList.class);
Dataset<R> result = readPath(spark, inputPath, resultClazz);
result
.joinWith(
possibleUpdates,
result.col("id").equalTo(possibleUpdates.col("resultId")),
"left_outer")
.map(contextUpdaterFn(), Encoders.bean(resultClazz))
.write()
.mode(SaveMode.Overwrite)
.option("compression", "gzip")
.json(outputPath);
result
.joinWith(
possibleUpdates,
result.col("id").equalTo(possibleUpdates.col("resultId")),
"left_outer")
.map(contextUpdaterFn(), Encoders.bean(resultClazz))
.write()
.mode(SaveMode.Overwrite)
.option("compression", "gzip")
.json(outputPath);
}
}
private static <R extends Result> MapFunction<Tuple2<R, ResultCommunityList>, R> contextUpdaterFn() {
return value -> {
R ret = value._1();
Optional<ResultCommunityList> rcl = Optional.ofNullable(value._2());
if (rcl.isPresent()) {
Set<String> contexts = new HashSet<>();
ret.getContext().forEach(c -> contexts.add(c.getId()));
rcl
.get()
.getCommunityList()
.stream()
.forEach(
c -> {
if (!contexts.contains(c)) {
Context newContext = new Context();
newContext.setId(c);
newContext
.setDataInfo(
Arrays
.asList(
getDataInfo(
PROPAGATION_DATA_INFO_TYPE,
PROPAGATION_RESULT_COMMUNITY_SEMREL_CLASS_ID,
PROPAGATION_RESULT_COMMUNITY_SEMREL_CLASS_NAME,
ModelConstants.DNET_PROVENANCE_ACTIONS)));
ret.getContext().add(newContext);
}
private static <R extends Result> MapFunction<Tuple2<R, ResultCommunityList>, R> contextUpdaterFn() {
return value -> {
R ret = value._1();
Optional<ResultCommunityList> rcl = Optional.ofNullable(value._2());
if (rcl.isPresent()) {
Set<String> contexts = new HashSet<>();
ret.getContext().forEach(c -> contexts.add(c.getId()));
rcl
.get()
.getCommunityList()
.stream()
.forEach(
c -> {
if (!contexts.contains(c)) {
Context newContext = new Context();
newContext.setId(c);
newContext
.setDataInfo(
Arrays
.asList(
getDataInfo(
PROPAGATION_DATA_INFO_TYPE,
PROPAGATION_RESULT_COMMUNITY_SEMREL_CLASS_ID,
PROPAGATION_RESULT_COMMUNITY_SEMREL_CLASS_NAME,
ModelConstants.DNET_PROVENANCE_ACTIONS)));
ret.getContext().add(newContext);
}
});
});
}
}
return ret;
};
}
return ret;
};
}
}

View File

@ -13,6 +13,7 @@ import org.apache.commons.io.IOUtils;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.api.java.function.FilterFunction;
import org.apache.spark.sql.Encoders;
import org.apache.spark.sql.Row;
import org.apache.spark.sql.SparkSession;
@ -31,18 +32,26 @@ public class BulkTagJobTest {
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
public static final String pathMap = "{ \"author\" : \"$['author'][*]['fullname']\","
+ " \"title\" : \"$['title'][*]['value']\","
+ " \"orcid\" : \"$['author'][*]['pid'][*][?(@['key']=='ORCID')]['value']\","
+ " \"contributor\" : \"$['contributor'][*]['value']\","
+ " \"description\" : \"$['description'][*]['value']\", "
+ " \"subject\" :\"$['subject'][*]['value']\" , " +
"\"fos\" : \"$['subject'][?(@['qualifier']['classid']=='FOS')].value\"," +
"\"sdg\" : \"$['subject'][?(@['qualifier']['classid']=='SDG')].value\"," +
"\"hostedby\" : \"$['instance'][*]['hostedby']['key']\" , " +
"\"collectedfrom\" : \"$['instance'][*]['collectedfrom']['key']\"," +
"\"publisher\":\"$['publisher'].value\"," +
"\"publicationyear\":\"$['dateofacceptance'].value\"} ";
public static final String pathMap = "{\"author\":{\"path\":\"$['author'][*]['fullname']\"}," +
" \"title\":{\"path\":\"$['title'][*]['value']\"}, " +
" \"orcid\":{\"path\":\"$['author'][*]['pid'][*][?(@['qualifier']['classid']=='orcid')]['value']\"} , " +
" \"orcid_pending\":{\"path\":\"$['author'][*]['pid'][*][?(@['qualifier']['classid']=='orcid_pending')]['value']\"} ,"
+
"\"contributor\" : {\"path\":\"$['contributor'][*]['value']\"}," +
" \"description\" : {\"path\":\"$['description'][*]['value']\"}," +
" \"subject\" :{\"path\":\"$['subject'][*]['value']\"}, " +
" \"fos\" : {\"path\":\"$['subject'][?(@['qualifier']['classid']=='FOS')].value\"} , " +
"\"sdg\" : {\"path\":\"$['subject'][?(@['qualifier']['classid']=='SDG')].value\"}," +
"\"journal\":{\"path\":\"$['journal'].name\"}," +
"\"hostedby\":{\"path\":\"$['instance'][*]['hostedby']['key']\"}," +
"\"collectedfrom\":{\"path\":\"$['instance'][*]['collectedfrom']['key']\"}," +
"\"publisher\":{\"path\":\"$['publisher'].value\"}," +
"\"publicationyear\":{\"path\":\"$['dateofacceptance'].value\", " +
" \"action\":{\"clazz\":\"eu.dnetlib.dhp.bulktag.actions.ExecSubstringAction\"," +
"\"method\":\"execSubstring\"," +
"\"params\":[" +
"{\"paramName\":\"From\", \"paramValue\":0}, " +
"{\"paramName\":\"To\",\"paramValue\":4}]}}}";
private static SparkSession spark;
@ -1600,4 +1609,94 @@ public class BulkTagJobTest {
Assertions.assertEquals(0, spark.sql(query).count());
}
@Test
void pubdateTest() throws Exception {
final String pathMap = BulkTagJobTest.pathMap;
SparkBulkTagJob
.main(
new String[] {
"-isSparkSessionManaged", Boolean.FALSE.toString(),
"-sourcePath",
getClass().getResource("/eu/dnetlib/dhp/bulktag/sample/dataset/publicationyear/").getPath(),
"-taggingConf",
IOUtils
.toString(
BulkTagJobTest.class
.getResourceAsStream(
"/eu/dnetlib/dhp/bulktag/communityconfiguration/tagging_conf_publicationdate.xml")),
"-outputPath", workingDir.toString() + "/",
"-pathMap", pathMap
});
final JavaSparkContext sc = JavaSparkContext.fromSparkContext(spark.sparkContext());
JavaRDD<Dataset> tmp = sc
.textFile(workingDir.toString() + "/dataset")
.map(item -> OBJECT_MAPPER.readValue(item, Dataset.class));
Assertions.assertEquals(10, tmp.count());
org.apache.spark.sql.Dataset<Dataset> verificationDataset = spark
.createDataset(tmp.rdd(), Encoders.bean(Dataset.class));
verificationDataset.createOrReplaceTempView("dataset");
String query = "select id, MyT.id community, MyD.provenanceaction.classid "
+ "from dataset "
+ "lateral view explode(context) c as MyT "
+ "lateral view explode(MyT.datainfo) d as MyD "
+ "where MyD.inferenceprovenance = 'bulktagging'";
org.apache.spark.sql.Dataset<Row> queryResult = spark.sql(query);
queryResult.show(false);
Assertions.assertEquals(5, queryResult.count());
Assertions
.assertEquals(
1,
queryResult
.filter(
(FilterFunction<Row>) r -> r
.getAs("id")
.equals("50|od______3989::02dd5d2c222191b0b9bd4f33c8e96529"))
.count());
Assertions
.assertEquals(
1,
queryResult
.filter(
(FilterFunction<Row>) r -> r
.getAs("id")
.equals("50|od______3989::2f4f3c820c450bd08dac08d07cc82dcf"))
.count());
Assertions
.assertEquals(
1,
queryResult
.filter(
(FilterFunction<Row>) r -> r
.getAs("id")
.equals("50|od______3989::7fcbe3a03280663cddebfd3cb9203177"))
.count());
Assertions
.assertEquals(
1,
queryResult
.filter(
(FilterFunction<Row>) r -> r
.getAs("id")
.equals("50|od______3989::d791339867bec6d3eb2104deeb4e4961"))
.count());
Assertions
.assertEquals(
1,
queryResult
.filter(
(FilterFunction<Row>) r -> r
.getAs("id")
.equals("50|od______3989::d90d3a1f64ad264b5ebed8a35b280343"))
.count());
}
}

Some files were not shown because too many files have changed in this diff Show More