new implementation that exploits prepared information (the list of possible updates: resultId - possible list of orcid to be added

This commit is contained in:
Miriam Baglioni 2020-04-16 15:55:25 +02:00
parent f077f22f73
commit d6cd700a32
1 changed files with 97 additions and 313 deletions

View File

@ -9,131 +9,115 @@ import org.apache.hadoop.fs.FileSystem;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaPairRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.Encoders;
import org.apache.spark.sql.Row;
import org.apache.spark.sql.SparkSession;
import org.apache.spark.sql.*;
import org.apache.spark.sql.Dataset;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import scala.Tuple2;
import java.util.*;
import static eu.dnetlib.dhp.PropagationConstant.*;
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkHiveSession;
public class SparkOrcidToResultFromSemRelJob3 {
private static final Logger log = LoggerFactory.getLogger(SparkOrcidToResultFromSemRelJob3.class);
private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper();
public class SparkOrcidToResultFromSemRelJob2 {
public static void main(String[] args) throws Exception {
String jsonConfiguration = IOUtils.toString(SparkOrcidToResultFromSemRelJob3.class
.getResourceAsStream("/eu/dnetlib/dhp/orcidtoresultfromsemrel/input_orcidtoresult_parameters.json"));
final ArgumentApplicationParser parser = new ArgumentApplicationParser(
jsonConfiguration);
final ArgumentApplicationParser parser = new ArgumentApplicationParser(IOUtils.toString(SparkOrcidToResultFromSemRelJob2.class.getResourceAsStream("/eu/dnetlib/dhp/orcidtoresultfromremrel/input_orcidtoresult_parameters.json")));
parser.parseArgument(args);
Boolean isSparkSessionManaged = isSparkSessionManaged(parser);
log.info("isSparkSessionManaged: {}", isSparkSessionManaged);
final String inputPath = parser.get("sourcePath");
log.info("inputPath: {}", inputPath);
final String outputPath = parser.get("outputPath");
log.info("outputPath: {}", outputPath);
final String possibleUpdates = parser.get("possibleUpdatesPath");
log.info("possibleUpdatesPath: {}", possibleUpdates);
final String resultClassName = parser.get("resultTableName");
log.info("resultTableName: {}", resultClassName);
final String resultType = resultClassName.substring(resultClassName.lastIndexOf(".") + 1).toLowerCase();
log.info("resultType: {}", resultType);
final Boolean saveGraph = Optional
.ofNullable(parser.get("saveGraph"))
.map(Boolean::valueOf)
.orElse(Boolean.TRUE);
log.info("saveGraph: {}", saveGraph);
Class<? extends Result> resultClazz = (Class<? extends Result>) Class.forName(resultClassName);
SparkConf conf = new SparkConf();
conf.set("hive.metastore.uris", parser.get("hive_metastore_uris"));
final SparkSession spark = SparkSession
.builder()
.appName(SparkOrcidToResultFromSemRelJob2.class.getSimpleName())
.master(parser.get("master"))
.config(conf)
.enableHiveSupport()
.getOrCreate();
final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext());
final String inputPath = parser.get("sourcePath");
final String outputPath = "/tmp/provision/propagation/orcidtoresult";
final List<String> allowedsemrel = Arrays.asList(parser.get("allowedsemrels").split(";"));
boolean writeUpdate = TRUE.equals(parser.get("writeUpdate"));
boolean saveGraph = TRUE.equals(parser.get("saveGraph"));
createOutputDirs(outputPath, FileSystem.get(spark.sparkContext().hadoopConfiguration()));
org.apache.spark.sql.Dataset<Relation> relation = spark.createDataset(sc.textFile(inputPath + "/relation")
.map(item -> new ObjectMapper().readValue(item, Relation.class)).rdd(), Encoders.bean(Relation.class));
org.apache.spark.sql.Dataset<Dataset> dataset = spark.createDataset(sc.textFile(inputPath + "/dataset")
.map(item -> new ObjectMapper().readValue(item, eu.dnetlib.dhp.schema.oaf.Dataset.class)).rdd(),
Encoders.bean(eu.dnetlib.dhp.schema.oaf.Dataset.class));
org.apache.spark.sql.Dataset<OtherResearchProduct> other = spark.createDataset(sc.textFile(inputPath + "/otherresearchproduct")
.map(item -> new ObjectMapper().readValue(item, eu.dnetlib.dhp.schema.oaf.OtherResearchProduct.class)).rdd(),
Encoders.bean(eu.dnetlib.dhp.schema.oaf.OtherResearchProduct.class));
org.apache.spark.sql.Dataset<Software> software = spark.createDataset(sc.textFile(inputPath + "/software")
.map(item -> new ObjectMapper().readValue(item, eu.dnetlib.dhp.schema.oaf.Software.class)).rdd(),
Encoders.bean(eu.dnetlib.dhp.schema.oaf.Software.class));
org.apache.spark.sql.Dataset<Publication> publication = spark.createDataset(sc.textFile(inputPath + "/publication")
.map(item -> new ObjectMapper().readValue(item, eu.dnetlib.dhp.schema.oaf.Publication.class)).rdd(),
Encoders.bean(eu.dnetlib.dhp.schema.oaf.Publication.class));
relation.createOrReplaceTempView("relation");
String query = "Select source, target " +
"from relation " +
"where datainfo.deletedbyinference = false " + getConstraintList(" relclass = '" , allowedsemrel);
org.apache.spark.sql.Dataset<Row> result_result = spark.sql(query);
publication.createOrReplaceTempView("publication");
org.apache.spark.sql.Dataset<ResultWithOrcid> pubs_with_orcid = getResultWithOrcid("publication", spark)
.as(Encoders.bean(ResultWithOrcid.class));
dataset.createOrReplaceTempView("dataset");
org.apache.spark.sql.Dataset<ResultWithOrcid> dats_with_orcid = getResultWithOrcid("dataset", spark)
.as(Encoders.bean(ResultWithOrcid.class));
other.createOrReplaceTempView("orp");
org.apache.spark.sql.Dataset<ResultWithOrcid> orp_with_orcid = getResultWithOrcid("orp", spark)
.as(Encoders.bean(ResultWithOrcid.class));
dataset.createOrReplaceTempView("software");
org.apache.spark.sql.Dataset<ResultWithOrcid> software_with_orcid = getResultWithOrcid("software", spark)
.as(Encoders.bean(ResultWithOrcid.class));
//get the results having at least one author pid we are interested in
//target of the relation from at least one source with orcid.
//the set of authors contains all those that have orcid and are related to target
//from any source with allowed semantic relationship
JavaPairRDD<String, List<AutoritativeAuthor>> target_authorlist_from_pubs = getTargetAutoritativeAuthorList(pubs_with_orcid);
JavaPairRDD<String, List<AutoritativeAuthor>> target_authorlist_from_dats = getTargetAutoritativeAuthorList(dats_with_orcid);
JavaPairRDD<String, List<AutoritativeAuthor>> target_authorlist_from_orp = getTargetAutoritativeAuthorList(orp_with_orcid);
JavaPairRDD<String, List<AutoritativeAuthor>> target_authorlist_from_sw = getTargetAutoritativeAuthorList(software_with_orcid);
if(writeUpdate){
target_authorlist_from_dats.map(r -> new ObjectMapper().writeValueAsString(r))
.saveAsTextFile(outputPath + "/" + "update_dats");
target_authorlist_from_pubs.map(r -> new ObjectMapper().writeValueAsString(r))
.saveAsTextFile(outputPath + "/" + "update_pubs");
target_authorlist_from_orp.map(r -> new ObjectMapper().writeValueAsString(r))
.saveAsTextFile(outputPath + "/" + "update_orp");
target_authorlist_from_sw.map(r -> new ObjectMapper().writeValueAsString(r))
.saveAsTextFile(outputPath + "/" + "update_sw");
}
if(saveGraph){
sc.textFile(inputPath + "/publication")
.map(item -> new ObjectMapper().readValue(item, eu.dnetlib.dhp.schema.oaf.Publication.class))
.mapToPair(p -> new Tuple2<>(p.getId(),p))
.leftOuterJoin(target_authorlist_from_pubs)
.map(c -> {
Result r = c._2()._1();
if(!c._2()._2().isPresent()){
return r;
}
List<eu.dnetlib.dhp.schema.oaf.Author> toenrich_author = r.getAuthor();
List<AutoritativeAuthor> autoritativeAuthors = c._2()._2().get();
for(eu.dnetlib.dhp.schema.oaf.Author author: toenrich_author){
if (!containsAllowedPid(author)){
enrichAuthor(author, autoritativeAuthors);
}
}
return r;
});
}
runWithSparkHiveSession(conf, isSparkSessionManaged,
spark -> {
if(isTest(parser)) {
removeOutputDir(spark, outputPath);
}
if(saveGraph)
execPropagation(spark, possibleUpdates, inputPath, outputPath, resultClazz);
});
}
private static void enrichAuthor(eu.dnetlib.dhp.schema.oaf.Author a, List<AutoritativeAuthor> au){
private static <R extends Result> void execPropagation(SparkSession spark, String possibleUpdatesPath, String inputPath,
String outputPath, Class<R> resultClazz ) {
//read possible updates (resultId and list of possible orcid to add
Dataset<ResultOrcidList> possible_updates = readAssocResultOrcidList(spark, possibleUpdatesPath);
//read the result we have been considering
Dataset<R> result = readPathEntity(spark, inputPath, resultClazz);
//make join result left_outer with possible updates
result.joinWith(possible_updates, result.col("id").equalTo(possible_updates.col("resultId")),
"left_outer")
.map(value -> {
R ret = value._1();
Optional<ResultOrcidList> rol = Optional.ofNullable(value._2());
if(rol.isPresent()) {
List<Author> toenrich_author = ret.getAuthor();
List<AutoritativeAuthor> autoritativeAuthors = rol.get().getAuthorList();
for(Author author: toenrich_author){
if (!containsAllowedPid(author)){
enrichAuthor(author, autoritativeAuthors);
}
}
}
return ret;
}, Encoders.bean(resultClazz))
.toJSON()
.write()
.mode(SaveMode.Overwrite)
.option("compression","gzip")
.text(outputPath);
}
private static Dataset<ResultOrcidList> readAssocResultOrcidList(SparkSession spark, String relationPath) {
return spark
.read()
.textFile(relationPath)
.map(value -> OBJECT_MAPPER.readValue(value, ResultOrcidList.class), Encoders.bean(ResultOrcidList.class));
}
private static void enrichAuthor(Author a, List<AutoritativeAuthor> au){
for (AutoritativeAuthor aa: au){
if(enrichAuthor(aa, a)){
return;
@ -142,77 +126,9 @@ public class SparkOrcidToResultFromSemRelJob2 {
}
// private static JavaPairRDD<String, List<AutoritativeAuthor>> getTargetAutoritativeAuthorList(org.apache.spark.sql.Dataset<Row> result_result, org.apache.spark.sql.Dataset<ResultWithOrcid> pubs_with_orcid) {
// return pubs_with_orcid
// .toJavaRDD()
// .mapToPair(p -> new Tuple2<>(p.getId(), p.getAuthorList()))
// .join(result_result.toJavaRDD().mapToPair(rel -> new Tuple2<>(rel.getString(0), rel.getString(1))))
// .mapToPair(c -> new Tuple2<>(c._2._2(), c._2()._1()))
// .reduceByKey((a, b) -> {
// if(a == null){
// return b;
// }
// if(b==null){
// return a;
// }
//
// Set<String> authSet = new HashSet<>();
// a.stream().forEach(au -> authSet.add(au.getOrcid()));
//
// b.stream().forEach(au -> {
// if (!authSet.contains(au.getOrcid())) {
// a.add(au);
// }
// }
// );
// return a;
// });
// }
private static JavaPairRDD<String, List<AutoritativeAuthor>> getTargetAutoritativeAuthorList( org.apache.spark.sql.Dataset<ResultWithOrcid> pubs_with_orcid) {
return pubs_with_orcid
.toJavaRDD()
.mapToPair(p -> new Tuple2<>(p.getId(), p.getAuthorList()))
.reduceByKey((a, b) -> {
if(a == null){
return b;
}
if(b==null){
return a;
}
Set<String> authSet = new HashSet<>();
a.stream().forEach(au -> authSet.add(au.getOrcid()));
b.stream().forEach(au -> {
if (!authSet.contains(au.getOrcid())) {
a.add(au);
}
}
);
return a;
});
}
private static org.apache.spark.sql.Dataset<Row> getResultWithOrcid(String table, SparkSession spark){
String query = " select target, author " +
" from (select id, collect_set(named_struct('name', name, 'surname', surname, 'fullname', fullname, 'orcid', orcid)) author " +
" from ( " +
" select id, MyT.fullname, MyT.name, MyT.surname, MyP.value orcid " +
" from " + table +
" lateral view explode (author) a as MyT " +
" lateral view explode (MyT.pid) p as MyP " +
" where MyP.qualifier.classid = 'ORCID') tmp " +
" group by id) r_t " +
" join (" +
" select source, target " +
" from relation " +
" where datainfo.deletedbyinference = false and (relclass = 'isSupplementedBy' or relclass = 'isSupplementTo') rel_rel " +
" on source = id";
return spark.sql(query);
}
private static boolean enrichAuthor(AutoritativeAuthor autoritative_author, eu.dnetlib.dhp.schema.oaf.Author author) {
private static boolean enrichAuthor(AutoritativeAuthor autoritative_author, Author author) {
boolean toaddpid = false;
if (StringUtils.isNoneEmpty(autoritative_author.getSurname())) {
@ -252,100 +168,9 @@ private static JavaPairRDD<String, List<AutoritativeAuthor>> getTargetAutoritati
}
// private static List<Author> enrichAuthors(List<Author> autoritative_authors, List<Author> to_enrich_authors, boolean filter){
//// List<Author> autoritative_authors = p._2()._2().get().getAuthors();
//// List<Author> to_enrich_authors = r.getAuthor();
//
// return to_enrich_authors
// .stream()
// .map(a -> {
// if (filter) {
// if (containsAllowedPid(a)) {
// return a;
// }
// }
//
// List<Author> lst = autoritative_authors.stream()
// .map(aa -> enrichAuthor(aa, a)).filter(au -> !(au == null)).collect(Collectors.toList());
// if (lst.size() == 0) {
// return a;
// }
// return lst.get(0);//Each author can be enriched at most once. It cannot be the same as many different people
//
// }).collect(Collectors.toList());
// }
//
// private static void writeResult(JavaPairRDD<String, Result> results, JavaPairRDD<String, TypedRow> toupdateresult,
// String outputPath, String type) {
//
// results.join(toupdateresult)
// .map(p -> {
// Result r = p._2()._1();
//
// List<Author> autoritative_authors = p._2()._2().getAuthors();
// List<eu.dnetlib.dhp.schema.oaf.Author> to_enrich_authors = r.getAuthor();
//
// r.setAuthor(enrichAutors(autoritative_authors, to_enrich_authors, false));
//// .stream()
//// .map(a -> {
//// if(filter) {
//// if (containsAllowedPid(a)) {
//// return a;
//// }
//// }
////
//// List<Author> lst = autoritative_authors.stream()
//// .map(aa -> enrichAuthor(aa, a)).filter(au -> !(au == null)).collect(Collectors.toList());
//// if(lst.size() == 0){
//// return a;
//// }
//// return lst.get(0);//Each author can be enriched at most once. It cannot be the same as many different people
////
//// }).collect(Collectors.toList()));
//
// return r;
// })
// .map(p -> new ObjectMapper().writeValueAsString(p))
// .saveAsTextFile(outputPath + "/" + type + "_update");
// }
// private static void updateResult(JavaPairRDD<String, Result> results, JavaPairRDD<String, TypedRow> toupdateresult,
// String outputPath, String type) {
// results.leftOuterJoin(toupdateresult)
// .map(p -> {
// Result r = p._2()._1();
// if (p._2()._2().isPresent()){
// List<AutoritativeAuthor> autoritative_authors = p._2()._2().get().getAuthors();
// List<eu.dnetlib.dhp.schema.oaf.Author> to_enrich_authors = r.getAuthor();
//
// r.setAuthor(enrichAutors(autoritative_authors, to_enrich_authors, true));
//// .stream()
//// .map(a -> {
//// if(filter) {
//// if (containsAllowedPid(a)) {
//// return a;
//// }
//// }
////
//// List<Author> lst = autoritative_authors.stream()
//// .map(aa -> enrichAuthor(aa, a)).filter(au -> !(au == null)).collect(Collectors.toList());
//// if(lst.size() == 0){
//// return a;
//// }
//// return lst.get(0);//Each author can be enriched at most once. It cannot be the same as many different people
////
//// }).collect(Collectors.toList()));
// }
// return r;
// })
// .map(p -> new ObjectMapper().writeValueAsString(p))
// .saveAsTextFile(outputPath+"/"+type);
// }
private static boolean containsAllowedPid(eu.dnetlib.dhp.schema.oaf.Author a) {
private static boolean containsAllowedPid(Author a) {
for (StructuredProperty pid : a.getPid()) {
if (PROPAGATION_AUTHOR_PID.equals(pid.getQualifier().getClassid())) {
return true;
@ -355,44 +180,3 @@ private static JavaPairRDD<String, List<AutoritativeAuthor>> getTargetAutoritati
}
}
/*private ResultProtos.Result.Metadata.Builder searchMatch(List<FieldTypeProtos.Author> author_list){
ResultProtos.Result.Metadata.Builder metadataBuilder = ResultProtos.Result.Metadata.newBuilder();
boolean updated = false;
for (FieldTypeProtos.Author a: author_list){
FieldTypeProtos.Author.Builder author = searchAuthor(a, autoritative_authors);
if(author != null){
updated = true;
metadataBuilder.addAuthor(author);
}else{
metadataBuilder.addAuthor(FieldTypeProtos.Author.newBuilder(a));
}
}
if(updated)
return metadataBuilder;
return null;
}
private FieldTypeProtos.Author.Builder searchAuthor(FieldTypeProtos.Author a, List<FieldTypeProtos.Author> author_list){
if(containsOrcid(a.getPidList()))
return null;
for(FieldTypeProtos.Author autoritative_author : author_list) {
if (equals(autoritative_author, a)) {
if(!containsOrcid(a.getPidList()))
return update(a, autoritative_author);
}
}
return null;
}
private boolean containsOrcid(List<FieldTypeProtos.KeyValue> pidList){
if(pidList == null)
return false;
return pidList
.stream()
.filter(kv -> kv.getKey().equals(PropagationConstants.AUTHOR_PID))
.collect(Collectors.toList()).size() > 0;
}
*/