Merge remote-tracking branch 'origin/stable_ids' into stable_ids

This commit is contained in:
Sandro La Bruzzo 2021-04-16 09:14:26 +02:00
commit 644aa8f40c
11 changed files with 192 additions and 70 deletions

View File

@ -57,7 +57,7 @@ public class IdentifierFactory implements Serializable {
} }
public static List<StructuredProperty> getPids(List<StructuredProperty> pid, KeyValue collectedFrom) { public static List<StructuredProperty> getPids(List<StructuredProperty> pid, KeyValue collectedFrom) {
return pidFromInstance(pid, collectedFrom).distinct().collect(Collectors.toList()); return pidFromInstance(pid, collectedFrom, true).distinct().collect(Collectors.toList());
} }
public static <T extends Result> String createDOIBoostIdentifier(T entity) { public static <T extends Result> String createDOIBoostIdentifier(T entity) {
@ -104,7 +104,7 @@ public class IdentifierFactory implements Serializable {
checkArgument(StringUtils.isNoneBlank(entity.getId()), "missing entity identifier"); checkArgument(StringUtils.isNoneBlank(entity.getId()), "missing entity identifier");
final Map<String, List<StructuredProperty>> pids = extractPids(entity); final Map<String, Set<StructuredProperty>> pids = extractPids(entity);
return pids return pids
.values() .values()
@ -125,7 +125,7 @@ public class IdentifierFactory implements Serializable {
.orElseGet(entity::getId); .orElseGet(entity::getId);
} }
private static <T extends OafEntity> Map<String, List<StructuredProperty>> extractPids(T entity) { private static <T extends OafEntity> Map<String, Set<StructuredProperty>> extractPids(T entity) {
if (entity instanceof Result) { if (entity instanceof Result) {
return Optional return Optional
.ofNullable(((Result) entity).getInstance()) .ofNullable(((Result) entity).getInstance())
@ -142,23 +142,24 @@ public class IdentifierFactory implements Serializable {
Collectors Collectors
.groupingBy( .groupingBy(
p -> p.getQualifier().getClassid(), p -> p.getQualifier().getClassid(),
Collectors.mapping(p -> p, Collectors.toList()))); Collectors.mapping(p -> p, Collectors.toCollection(HashSet::new))));
} }
} }
private static Map<String, List<StructuredProperty>> mapPids(List<Instance> instance) { private static Map<String, Set<StructuredProperty>> mapPids(List<Instance> instance) {
return instance return instance
.stream() .stream()
.map(i -> pidFromInstance(i.getPid(), i.getCollectedfrom())) .map(i -> pidFromInstance(i.getPid(), i.getCollectedfrom(), false))
.flatMap(Function.identity()) .flatMap(Function.identity())
.collect( .collect(
Collectors Collectors
.groupingBy( .groupingBy(
p -> p.getQualifier().getClassid(), p -> p.getQualifier().getClassid(),
Collectors.mapping(p -> p, Collectors.toList()))); Collectors.mapping(p -> p, Collectors.toCollection(HashSet::new))));
} }
private static Stream<StructuredProperty> pidFromInstance(List<StructuredProperty> pid, KeyValue collectedFrom) { private static Stream<StructuredProperty> pidFromInstance(List<StructuredProperty> pid, KeyValue collectedFrom,
boolean mapHandles) {
return Optional return Optional
.ofNullable(pid) .ofNullable(pid)
.map( .map(
@ -167,16 +168,16 @@ public class IdentifierFactory implements Serializable {
// filter away PIDs provided by a DS that is not considered an authority for the // filter away PIDs provided by a DS that is not considered an authority for the
// given PID Type // given PID Type
.filter(p -> { .filter(p -> {
return shouldFilterPid(collectedFrom, p); return shouldFilterPid(collectedFrom, p, mapHandles);
}) })
.map(CleaningFunctions::normalizePidValue) .map(CleaningFunctions::normalizePidValue)
.filter(CleaningFunctions::pidFilter)) .filter(CleaningFunctions::pidFilter))
.orElse(Stream.empty()); .orElse(Stream.empty());
} }
private static boolean shouldFilterPid(KeyValue collectedFrom, StructuredProperty p) { private static boolean shouldFilterPid(KeyValue collectedFrom, StructuredProperty p, boolean mapHandles) {
final PidType pType = PidType.tryValueOf(p.getQualifier().getClassid()); final PidType pType = PidType.tryValueOf(p.getQualifier().getClassid());
return pType.equals(PidType.handle) || Optional.ofNullable(collectedFrom).isPresent() && return (mapHandles && pType.equals(PidType.handle)) || Optional.ofNullable(collectedFrom).isPresent() &&
Optional Optional
.ofNullable(PID_AUTHORITY.get(pType)) .ofNullable(PID_AUTHORITY.get(pType))
.map(authorities -> { .map(authorities -> {

View File

@ -31,6 +31,9 @@ public class IdentifierFactoryTest {
verifyIdentifier( verifyIdentifier(
"publication_doi3.json", "50|pmc_________::94e4cb08c93f8733b48e2445d04002ac", true); "publication_doi3.json", "50|pmc_________::94e4cb08c93f8733b48e2445d04002ac", true);
verifyIdentifier(
"publication_doi4.json", "50|od______2852::38861c44e6052a8d49f59a4c39ba5e66", true);
verifyIdentifier( verifyIdentifier(
"publication_pmc1.json", "50|DansKnawCris::0829b5191605bdbea36d6502b8c1ce1f", true); "publication_pmc1.json", "50|DansKnawCris::0829b5191605bdbea36d6502b8c1ce1f", true);

View File

@ -0,0 +1,37 @@
{
"id": "50|od______2852::38861c44e6052a8d49f59a4c39ba5e66",
"instance": [
{
"collectedfrom": {
"key": "10|openaire____::1234",
"value": "Zenodo"
},
"pid": [
{
"qualifier": {"classid": "doi"},
"value": "10.1016/j.cmet.2010.03.013"
},
{
"qualifier": {"classid": "handle"},
"value": "11012/83840"
}
]
},
{
"collectedfrom": {
"key": "10|opendoar____::2852",
"value": "Digital library of Brno University of Technology"
},
"pid": [
{
"qualifier": {"classid": "pmc"},
"value": "21459329"
},
{
"qualifier": {"classid": "handle"},
"value": "11012/83840"
}
]
}
]
}

View File

@ -36,7 +36,6 @@ public class ModelConstants {
public static final String DNET_COUNTRY_TYPE = "dnet:countries"; public static final String DNET_COUNTRY_TYPE = "dnet:countries";
public static final String DNET_REVIEW_LEVELS = "dnet:review_levels"; public static final String DNET_REVIEW_LEVELS = "dnet:review_levels";
public static final String DNET_PROGRAMMING_LANGUAGES = "dnet:programming_languages"; public static final String DNET_PROGRAMMING_LANGUAGES = "dnet:programming_languages";
public static final String DNET_PROVENANCEACTIONS = "dnet:provenanceActions";
public static final String DNET_EXTERNAL_REF_TYPES = "dnet:externalReference_typologies"; public static final String DNET_EXTERNAL_REF_TYPES = "dnet:externalReference_typologies";
public static final String SYSIMPORT_CROSSWALK_REPOSITORY = "sysimport:crosswalk:repository"; public static final String SYSIMPORT_CROSSWALK_REPOSITORY = "sysimport:crosswalk:repository";
@ -50,7 +49,7 @@ public class ModelConstants {
public static final String PROVENANCE_DEDUP = "sysimport:dedup"; public static final String PROVENANCE_DEDUP = "sysimport:dedup";
public static final Qualifier PROVENANCE_ACTION_SET_QUALIFIER = qualifier( public static final Qualifier PROVENANCE_ACTION_SET_QUALIFIER = qualifier(
SYSIMPORT_ACTIONSET, SYSIMPORT_ACTIONSET, DNET_PROVENANCEACTIONS, DNET_PROVENANCEACTIONS); SYSIMPORT_ACTIONSET, SYSIMPORT_ACTIONSET, DNET_PROVENANCE_ACTIONS, DNET_PROVENANCE_ACTIONS);
public static final String DATASET_RESULTTYPE_CLASSID = "dataset"; public static final String DATASET_RESULTTYPE_CLASSID = "dataset";
public static final String PUBLICATION_RESULTTYPE_CLASSID = "publication"; public static final String PUBLICATION_RESULTTYPE_CLASSID = "publication";

View File

@ -82,16 +82,12 @@ public class DedupRecordFactory {
final Collection<String> dates = Lists.newArrayList(); final Collection<String> dates = Lists.newArrayList();
final List<List<Author>> authors = Lists.newArrayList(); final List<List<Author>> authors = Lists.newArrayList();
final List<Identifier<T>> bestPids = Lists.newArrayList(); // best pids list
entities entities
.forEachRemaining( .forEachRemaining(
t -> { t -> {
T duplicate = t._2(); T duplicate = t._2();
// prepare the list of pids to be used for the id generation
bestPids.add(Identifier.newInstance(duplicate));
entity.mergeFrom(duplicate); entity.mergeFrom(duplicate);
if (ModelSupport.isSubClass(duplicate, Result.class)) { if (ModelSupport.isSubClass(duplicate, Result.class)) {
Result r1 = (Result) duplicate; Result r1 = (Result) duplicate;
@ -109,7 +105,7 @@ public class DedupRecordFactory {
((Result) entity).setAuthor(AuthorMerger.merge(authors)); ((Result) entity).setAuthor(AuthorMerger.merge(authors));
} }
entity.setId(IdGenerator.generate(bestPids, id)); entity.setId(id);
entity.setLastupdatetimestamp(ts); entity.setLastupdatetimestamp(ts);
entity.setDataInfo(dataInfo); entity.setDataInfo(dataInfo);

View File

@ -1,6 +1,9 @@
package eu.dnetlib.dhp.oa.dedup; package eu.dnetlib.dhp.oa.dedup;
import static eu.dnetlib.dhp.schema.common.ModelConstants.DNET_PROVENANCE_ACTIONS;
import static eu.dnetlib.dhp.schema.common.ModelConstants.PROVENANCE_DEDUP;
import java.io.IOException; import java.io.IOException;
import org.apache.commons.io.IOUtils; import org.apache.commons.io.IOUtils;
@ -27,8 +30,6 @@ public class SparkCreateDedupRecord extends AbstractSparkAction {
private static final Logger log = LoggerFactory.getLogger(SparkCreateDedupRecord.class); private static final Logger log = LoggerFactory.getLogger(SparkCreateDedupRecord.class);
public static final String ROOT_TRUST = "0.8"; public static final String ROOT_TRUST = "0.8";
public static final String PROVENANCE_ACTION_CLASS = "sysimport:dedup";
public static final String PROVENANCE_ACTIONS = "dnet:provenanceActions";
public SparkCreateDedupRecord(ArgumentApplicationParser parser, SparkSession spark) { public SparkCreateDedupRecord(ArgumentApplicationParser parser, SparkSession spark) {
super(parser, spark); super(parser, spark);
@ -94,10 +95,10 @@ public class SparkCreateDedupRecord extends AbstractSparkAction {
info.setTrust(ROOT_TRUST); info.setTrust(ROOT_TRUST);
info.setInferenceprovenance(dedupConf.getWf().getConfigurationId()); info.setInferenceprovenance(dedupConf.getWf().getConfigurationId());
Qualifier provenance = new Qualifier(); Qualifier provenance = new Qualifier();
provenance.setClassid(PROVENANCE_ACTION_CLASS); provenance.setClassid(PROVENANCE_DEDUP);
provenance.setClassname(PROVENANCE_ACTION_CLASS); provenance.setClassname(PROVENANCE_DEDUP);
provenance.setSchemeid(PROVENANCE_ACTIONS); provenance.setSchemeid(DNET_PROVENANCE_ACTIONS);
provenance.setSchemename(PROVENANCE_ACTIONS); provenance.setSchemename(DNET_PROVENANCE_ACTIONS);
info.setProvenanceaction(provenance); info.setProvenanceaction(provenance);
return info; return info;
} }

View File

@ -1,18 +1,20 @@
package eu.dnetlib.dhp.oa.dedup; package eu.dnetlib.dhp.oa.dedup;
import static eu.dnetlib.dhp.schema.common.ModelConstants.DNET_PROVENANCE_ACTIONS;
import static eu.dnetlib.dhp.schema.common.ModelConstants.PROVENANCE_DEDUP;
import java.io.IOException; import java.io.IOException;
import java.util.ArrayList; import java.util.*;
import java.util.Iterator; import java.util.stream.Collectors;
import java.util.List;
import java.util.Optional;
import org.apache.commons.io.IOUtils; import org.apache.commons.io.IOUtils;
import org.apache.spark.SparkConf; import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaPairRDD; import org.apache.spark.api.java.JavaPairRDD;
import org.apache.spark.api.java.JavaSparkContext; import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.api.java.function.FlatMapFunction;
import org.apache.spark.api.java.function.MapFunction; import org.apache.spark.api.java.function.MapFunction;
import org.apache.spark.api.java.function.PairFunction; import org.apache.spark.api.java.function.MapGroupsFunction;
import org.apache.spark.graphx.Edge; import org.apache.spark.graphx.Edge;
import org.apache.spark.rdd.RDD; import org.apache.spark.rdd.RDD;
import org.apache.spark.sql.Dataset; import org.apache.spark.sql.Dataset;
@ -23,14 +25,18 @@ import org.dom4j.DocumentException;
import org.slf4j.Logger; import org.slf4j.Logger;
import org.slf4j.LoggerFactory; import org.slf4j.LoggerFactory;
import com.google.common.collect.Lists;
import com.google.common.hash.Hashing; import com.google.common.hash.Hashing;
import eu.dnetlib.dhp.application.ArgumentApplicationParser; import eu.dnetlib.dhp.application.ArgumentApplicationParser;
import eu.dnetlib.dhp.oa.dedup.graph.ConnectedComponent; import eu.dnetlib.dhp.oa.dedup.graph.ConnectedComponent;
import eu.dnetlib.dhp.oa.dedup.graph.GraphProcessor; import eu.dnetlib.dhp.oa.dedup.graph.GraphProcessor;
import eu.dnetlib.dhp.oa.dedup.model.Identifier;
import eu.dnetlib.dhp.schema.common.EntityType;
import eu.dnetlib.dhp.schema.common.ModelConstants; import eu.dnetlib.dhp.schema.common.ModelConstants;
import eu.dnetlib.dhp.schema.common.ModelSupport; import eu.dnetlib.dhp.schema.common.ModelSupport;
import eu.dnetlib.dhp.schema.oaf.DataInfo; import eu.dnetlib.dhp.schema.oaf.DataInfo;
import eu.dnetlib.dhp.schema.oaf.OafEntity;
import eu.dnetlib.dhp.schema.oaf.Qualifier; import eu.dnetlib.dhp.schema.oaf.Qualifier;
import eu.dnetlib.dhp.schema.oaf.Relation; import eu.dnetlib.dhp.schema.oaf.Relation;
import eu.dnetlib.dhp.utils.ISLookupClientFactory; import eu.dnetlib.dhp.utils.ISLookupClientFactory;
@ -42,9 +48,7 @@ import scala.Tuple2;
public class SparkCreateMergeRels extends AbstractSparkAction { public class SparkCreateMergeRels extends AbstractSparkAction {
public static final String PROVENANCE_ACTION_CLASS = "sysimport:dedup";
private static final Logger log = LoggerFactory.getLogger(SparkCreateMergeRels.class); private static final Logger log = LoggerFactory.getLogger(SparkCreateMergeRels.class);
public static final String DNET_PROVENANCE_ACTIONS = "dnet:provenanceActions";
public SparkCreateMergeRels(ArgumentApplicationParser parser, SparkSession spark) { public SparkCreateMergeRels(ArgumentApplicationParser parser, SparkSession spark) {
super(parser, spark); super(parser, spark);
@ -92,6 +96,7 @@ public class SparkCreateMergeRels extends AbstractSparkAction {
for (DedupConfig dedupConf : getConfigurations(isLookUpService, actionSetId)) { for (DedupConfig dedupConf : getConfigurations(isLookUpService, actionSetId)) {
final String subEntity = dedupConf.getWf().getSubEntityValue(); final String subEntity = dedupConf.getWf().getSubEntityValue();
final Class<OafEntity> clazz = ModelSupport.entityTypes.get(EntityType.valueOf(subEntity));
log.info("Creating mergerels for: '{}'", subEntity); log.info("Creating mergerels for: '{}'", subEntity);
@ -100,10 +105,8 @@ public class SparkCreateMergeRels extends AbstractSparkAction {
final String mergeRelPath = DedupUtility.createMergeRelPath(workingPath, actionSetId, subEntity); final String mergeRelPath = DedupUtility.createMergeRelPath(workingPath, actionSetId, subEntity);
final JavaPairRDD<Object, String> vertexes = sc // <hash(id), id>
.textFile(graphBasePath + "/" + subEntity) JavaPairRDD<Object, String> vertexes = createVertexes(sc, graphBasePath, subEntity, dedupConf);
.map(s -> MapDocumentUtil.getJPathString(dedupConf.getWf().getIdPath(), s))
.mapToPair((PairFunction<String, Object, String>) s -> new Tuple2<>(hash(s), s));
final RDD<Edge<String>> edgeRdd = spark final RDD<Edge<String>> edgeRdd = spark
.read() .read()
@ -113,14 +116,42 @@ public class SparkCreateMergeRels extends AbstractSparkAction {
.map(it -> new Edge<>(hash(it.getSource()), hash(it.getTarget()), it.getRelClass())) .map(it -> new Edge<>(hash(it.getSource()), hash(it.getTarget()), it.getRelClass()))
.rdd(); .rdd();
final Dataset<Relation> mergeRels = spark Dataset<Tuple2<String, String>> rawMergeRels = spark
.createDataset( .createDataset(
GraphProcessor GraphProcessor
.findCCs(vertexes.rdd(), edgeRdd, maxIterations, cut) .findCCs(vertexes.rdd(), edgeRdd, maxIterations, cut)
.toJavaRDD() .toJavaRDD()
.filter(k -> k.getDocIds().size() > 1) .filter(k -> k.getIds().size() > 1)
.flatMap(cc -> ccToMergeRel(cc, dedupConf)) .flatMap(this::ccToRels)
.rdd(), .rdd(),
Encoders.tuple(Encoders.STRING(), Encoders.STRING()));
Dataset<Tuple2<String, OafEntity>> entities = spark
.read()
.textFile(DedupUtility.createEntityPath(graphBasePath, subEntity))
.map(
(MapFunction<String, Tuple2<String, OafEntity>>) it -> {
OafEntity entity = OBJECT_MAPPER.readValue(it, clazz);
return new Tuple2<>(entity.getId(), entity);
},
Encoders.tuple(Encoders.STRING(), Encoders.kryo(clazz)));
Dataset<Relation> mergeRels = rawMergeRels
.joinWith(entities, rawMergeRels.col("_2").equalTo(entities.col("_1")), "inner")
// <tmp_source,target>,<target,entity>
.map(
(MapFunction<Tuple2<Tuple2<String, String>, Tuple2<String, OafEntity>>, Tuple2<String, OafEntity>>) value -> new Tuple2<>(
value._1()._1(), value._2()._2()),
Encoders.tuple(Encoders.STRING(), Encoders.kryo(clazz)))
// <tmp_source,entity>
.groupByKey(
(MapFunction<Tuple2<String, OafEntity>, String>) Tuple2::_1, Encoders.STRING())
.mapGroups(
(MapGroupsFunction<String, Tuple2<String, OafEntity>, ConnectedComponent>) this::generateID,
Encoders.bean(ConnectedComponent.class))
// <root_id, list(target)>
.flatMap(
(FlatMapFunction<ConnectedComponent, Relation>) cc -> ccToMergeRel(cc, dedupConf),
Encoders.bean(Relation.class)); Encoders.bean(Relation.class));
mergeRels.write().mode(SaveMode.Append).parquet(mergeRelPath); mergeRels.write().mode(SaveMode.Append).parquet(mergeRelPath);
@ -128,9 +159,45 @@ public class SparkCreateMergeRels extends AbstractSparkAction {
} }
} }
public Iterator<Relation> ccToMergeRel(ConnectedComponent cc, DedupConfig dedupConf) { private <T extends OafEntity> ConnectedComponent generateID(String key, Iterator<Tuple2<String, T>> values) {
List<Identifier<T>> identifiers = Lists.newArrayList(values).stream().map(v -> {
T entity = v._2();
Identifier<T> identifier = Identifier.newInstance(entity);
return identifier;
}).collect(Collectors.toList());
String rootID = IdGenerator.generate(identifiers, key);
if (Objects.equals(rootID, key))
throw new IllegalStateException("generated default ID: " + rootID);
return new ConnectedComponent(rootID,
identifiers.stream().map(i -> i.getEntity().getId()).collect(Collectors.toSet()));
}
private JavaPairRDD<Object, String> createVertexes(JavaSparkContext sc, String graphBasePath, String subEntity,
DedupConfig dedupConf) {
return sc
.textFile(DedupUtility.createEntityPath(graphBasePath, subEntity))
.mapToPair(json -> {
String id = MapDocumentUtil.getJPathString(dedupConf.getWf().getIdPath(), json);
return new Tuple2<>(hash(id), id);
});
}
private Iterator<Tuple2<String, String>> ccToRels(ConnectedComponent cc) {
return cc return cc
.getDocIds() .getIds()
.stream()
.map(id -> new Tuple2<>(cc.getCcId(), id))
.iterator();
}
private Iterator<Relation> ccToMergeRel(ConnectedComponent cc, DedupConfig dedupConf) {
return cc
.getIds()
.stream() .stream()
.flatMap( .flatMap(
id -> { id -> {
@ -161,8 +228,8 @@ public class SparkCreateMergeRels extends AbstractSparkAction {
info.setInvisible(false); info.setInvisible(false);
info.setInferenceprovenance(dedupConf.getWf().getConfigurationId()); info.setInferenceprovenance(dedupConf.getWf().getConfigurationId());
Qualifier provenanceAction = new Qualifier(); Qualifier provenanceAction = new Qualifier();
provenanceAction.setClassid(PROVENANCE_ACTION_CLASS); provenanceAction.setClassid(PROVENANCE_DEDUP);
provenanceAction.setClassname(PROVENANCE_ACTION_CLASS); provenanceAction.setClassname(PROVENANCE_DEDUP);
provenanceAction.setSchemeid(DNET_PROVENANCE_ACTIONS); provenanceAction.setSchemeid(DNET_PROVENANCE_ACTIONS);
provenanceAction.setSchemename(DNET_PROVENANCE_ACTIONS); provenanceAction.setSchemename(DNET_PROVENANCE_ACTIONS);
info.setProvenanceaction(provenanceAction); info.setProvenanceaction(provenanceAction);

View File

@ -3,44 +3,66 @@ package eu.dnetlib.dhp.oa.dedup.graph;
import java.io.IOException; import java.io.IOException;
import java.io.Serializable; import java.io.Serializable;
import java.util.ArrayList;
import java.util.List;
import java.util.Objects;
import java.util.Set; import java.util.Set;
import java.util.stream.Collectors; import java.util.stream.Collectors;
import org.apache.commons.lang.StringUtils; import org.apache.commons.lang.StringUtils;
import org.apache.spark.api.java.function.MapFunction;
import org.codehaus.jackson.annotate.JsonIgnore; import org.codehaus.jackson.annotate.JsonIgnore;
import com.fasterxml.jackson.databind.DeserializationFeature;
import com.fasterxml.jackson.databind.ObjectMapper; import com.fasterxml.jackson.databind.ObjectMapper;
import com.google.common.collect.Lists;
import eu.dnetlib.dhp.oa.dedup.DedupUtility; import eu.dnetlib.dhp.oa.dedup.DedupUtility;
import eu.dnetlib.dhp.oa.dedup.IdGenerator;
import eu.dnetlib.dhp.oa.dedup.model.Identifier;
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.utils.DHPUtils; import eu.dnetlib.dhp.utils.DHPUtils;
import eu.dnetlib.pace.config.DedupConfig;
import eu.dnetlib.pace.util.MapDocumentUtil;
import eu.dnetlib.pace.util.PaceException; import eu.dnetlib.pace.util.PaceException;
public class ConnectedComponent implements Serializable { public class ConnectedComponent implements Serializable {
private Set<String> docIds;
private String ccId; private String ccId;
private Set<String> ids;
public ConnectedComponent(Set<String> docIds, final int cut) { private static final String CONNECTED_COMPONENT_ID_PREFIX = "connect_comp";
this.docIds = docIds;
createID(); public ConnectedComponent(Set<String> ids, final int cut) {
if (cut > 0 && docIds.size() > cut) { this.ids = ids;
this.docIds = docIds
this.ccId = createDefaultID();
if (cut > 0 && ids.size() > cut) {
this.ids = ids
.stream() .stream()
.filter(s -> !ccId.equalsIgnoreCase(s)) .filter(id -> !ccId.equalsIgnoreCase(id))
.limit(cut - 1) .limit(cut - 1)
.collect(Collectors.toSet()); .collect(Collectors.toSet());
this.docIds.add(ccId); // this.ids.add(ccId); ??
} }
} }
public String createID() { public ConnectedComponent(String ccId, Set<String> ids) {
if (docIds.size() > 1) { this.ccId = ccId;
this.ids = ids;
}
public String createDefaultID() {
if (ids.size() > 1) {
final String s = getMin(); final String s = getMin();
String prefix = s.split("\\|")[0]; String prefix = s.split("\\|")[0];
ccId = prefix + "|dedup_wf_001::" + DHPUtils.md5(s); ccId = prefix + "|" + CONNECTED_COMPONENT_ID_PREFIX + "::" + DHPUtils.md5(s);
return ccId; return ccId;
} else { } else {
return docIds.iterator().next(); return ids.iterator().next();
} }
} }
@ -49,15 +71,15 @@ public class ConnectedComponent implements Serializable {
final StringBuilder min = new StringBuilder(); final StringBuilder min = new StringBuilder();
docIds ids
.forEach( .forEach(
i -> { id -> {
if (StringUtils.isBlank(min.toString())) { if (StringUtils.isBlank(min.toString())) {
min.append(i); min.append(id);
} else { } else {
if (min.toString().compareTo(i) > 0) { if (min.toString().compareTo(id) > 0) {
min.setLength(0); min.setLength(0);
min.append(i); min.append(id);
} }
} }
}); });
@ -74,12 +96,12 @@ public class ConnectedComponent implements Serializable {
} }
} }
public Set<String> getDocIds() { public Set<String> getIds() {
return docIds; return ids;
} }
public void setDocIds(Set<String> docIds) { public void setIds(Set<String> ids) {
this.docIds = docIds; this.ids = ids;
} }
public String getCcId() { public String getCcId() {

View File

@ -2,7 +2,6 @@
package eu.dnetlib.dhp.oa.dedup.model; package eu.dnetlib.dhp.oa.dedup.model;
import java.io.Serializable; import java.io.Serializable;
import java.text.ParseException;
import java.text.SimpleDateFormat; import java.text.SimpleDateFormat;
import java.util.*; import java.util.*;
import java.util.stream.Collectors; import java.util.stream.Collectors;

View File

@ -55,7 +55,6 @@ public class SparkDedupTest implements Serializable {
private static String testOutputBasePath; private static String testOutputBasePath;
private static String testDedupGraphBasePath; private static String testDedupGraphBasePath;
private static final String testActionSetId = "test-orchestrator"; private static final String testActionSetId = "test-orchestrator";
private static String testDedupAssertionsBasePath;
@BeforeAll @BeforeAll
public static void cleanUp() throws IOException, URISyntaxException { public static void cleanUp() throws IOException, URISyntaxException {
@ -70,10 +69,6 @@ public class SparkDedupTest implements Serializable {
testDedupGraphBasePath = createTempDirectory(SparkDedupTest.class.getSimpleName() + "-") testDedupGraphBasePath = createTempDirectory(SparkDedupTest.class.getSimpleName() + "-")
.toAbsolutePath() .toAbsolutePath()
.toString(); .toString();
testDedupAssertionsBasePath = Paths
.get(SparkDedupTest.class.getResource("/eu/dnetlib/dhp/dedup/assertions").toURI())
.toFile()
.getAbsolutePath();
FileUtils.deleteDirectory(new File(testOutputBasePath)); FileUtils.deleteDirectory(new File(testOutputBasePath));
FileUtils.deleteDirectory(new File(testDedupGraphBasePath)); FileUtils.deleteDirectory(new File(testDedupGraphBasePath));
@ -353,6 +348,7 @@ public class SparkDedupTest implements Serializable {
assertEquals(288, sw_mergerel); assertEquals(288, sw_mergerel);
assertEquals(472, ds_mergerel); assertEquals(472, ds_mergerel);
assertEquals(718, orp_mergerel); assertEquals(718, orp_mergerel);
} }
@Test @Test
@ -538,7 +534,7 @@ public class SparkDedupTest implements Serializable {
long relations = jsc.textFile(testDedupGraphBasePath + "/relation").count(); long relations = jsc.textFile(testDedupGraphBasePath + "/relation").count();
assertEquals(4858, relations); assertEquals(4862, relations);
// check deletedbyinference // check deletedbyinference
final Dataset<Relation> mergeRels = spark final Dataset<Relation> mergeRels = spark

View File

@ -10,6 +10,7 @@ export SOURCE=$1
export SHADOW=$2 export SHADOW=$2
echo "Updating shadow database" echo "Updating shadow database"
impala-shell -q "invalidate metadata"
impala-shell -d ${SOURCE} -q "invalidate metadata" impala-shell -d ${SOURCE} -q "invalidate metadata"
impala-shell -d ${SOURCE} -q "show tables" --delimited | sed "s/^\(.*\)/compute stats ${SOURCE}.\1;/" | impala-shell -c -f - impala-shell -d ${SOURCE} -q "show tables" --delimited | sed "s/^\(.*\)/compute stats ${SOURCE}.\1;/" | impala-shell -c -f -
impala-shell -q "create database if not exists ${SHADOW}" impala-shell -q "create database if not exists ${SHADOW}"