From 03f7cb64028fc1f215d4c549bd16b46ea29f0e1b Mon Sep 17 00:00:00 2001 From: "miriam.baglioni" Date: Wed, 8 Apr 2020 18:08:41 +0200 Subject: [PATCH] new parametrized implementation for country propagation --- .../PrepareResultCountryAssociation.java | 68 ++-- .../SparkCountryPropagationJob2.java | 298 ++++++------------ 2 files changed, 133 insertions(+), 233 deletions(-) diff --git a/dhp-workflows/dhp-propagation/src/main/java/eu/dnetlib/dhp/countrypropagation/PrepareResultCountryAssociation.java b/dhp-workflows/dhp-propagation/src/main/java/eu/dnetlib/dhp/countrypropagation/PrepareResultCountryAssociation.java index 9572159ce5..b84c4afb54 100644 --- a/dhp-workflows/dhp-propagation/src/main/java/eu/dnetlib/dhp/countrypropagation/PrepareResultCountryAssociation.java +++ b/dhp-workflows/dhp-propagation/src/main/java/eu/dnetlib/dhp/countrypropagation/PrepareResultCountryAssociation.java @@ -14,14 +14,13 @@ import org.apache.spark.sql.SaveMode; import java.util.Arrays; import java.util.List; -import java.util.Optional; + import static eu.dnetlib.dhp.PropagationConstant.createOutputDirs; import static eu.dnetlib.dhp.PropagationConstant.getConstraintList; import org.slf4j.Logger; import org.slf4j.LoggerFactory; -import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession; /** * For the association of the country to the datasource @@ -38,53 +37,46 @@ public class PrepareResultCountryAssociation { public static void main(String[] args) throws Exception { String jsonConfiguration = IOUtils.toString(PrepareResultCountryAssociation.class - .getResourceAsStream("/eu/dnetlib/dhp/countrypropagation/input_countrypropagation_parameters.json")); + .getResourceAsStream("/eu/dnetlib/dhp/countrypropagation/input_prepare_dc_assoc.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); - String inputPath = parser.get("sourcePath"); log.info("inputPath: {}", inputPath); - - String graphTableClassName = parser.get("graphTableClassName"); - log.info("graphTableClassName: {}", graphTableClassName); - SparkConf conf = new SparkConf(); conf.set("hive.metastore.uris", parser.get("hive_metastore_uris")); - runWithSparkSession(conf, isSparkSessionManaged, - spark -> { - removeOutputDir(spark, outputPath); - joinRelationEntity(spark, inputRelationsPath, inputEntityPath, entityClazz, outputPath); - }); - - final SparkSession spark = SparkSession .builder() - .appName(PrepareResultCountryAssociation.class.getSimpleName()) + .appName(SparkCountryPropagationJob.class.getSimpleName()) .master(parser.get("master")) .config(conf) .enableHiveSupport() .getOrCreate(); - //todo add link to working dir + final String outputPath = "/tmp/provision/propagation/countrytoresultfrominstitutionalrepositories"; + createOutputDirs(outputPath, FileSystem.get(spark.sparkContext().hadoopConfiguration())); + prepareDatasourceCountryAssociation(spark, + Arrays.asList(parser.get("whitelist").split(";")), + Arrays.asList(parser.get("allowedtypes").split(";")), + inputPath, + outputPath); - List whitelist = Arrays.asList(parser.get("whitelist").split(";")); - List allowedtypes = Arrays.asList(parser.get("allowedtypes").split(";")); + } + + private static void prepareDatasourceCountryAssociation(SparkSession spark, + List whitelist, + List allowedtypes, + String inputPath, + String outputPath) { String whitelisted = ""; for (String i : whitelist){ whitelisted += " OR id = '" + i + "'"; @@ -93,13 +85,13 @@ public class PrepareResultCountryAssociation { Dataset datasource = spark.createDataset(sc.textFile(inputPath + "/datasource") - .map(item -> new ObjectMapper().readValue(item, Datasource.class)).rdd(), Encoders.bean(Datasource.class)); + .map(item -> OBJECT_MAPPER.readValue(item, Datasource.class)).rdd(), Encoders.bean(Datasource.class)); Dataset relation = spark.createDataset(sc.textFile(inputPath + "/relation") - .map(item -> new ObjectMapper().readValue(item, Relation.class)).rdd(), Encoders.bean(Relation.class)); + .map(item -> OBJECT_MAPPER.readValue(item, Relation.class)).rdd(), Encoders.bean(Relation.class)); Dataset organization = spark.createDataset(sc.textFile(inputPath + "/organization") - .map(item -> new ObjectMapper().readValue(item, Organization.class)).rdd(), Encoders.bean(Organization.class)); + .map(item -> OBJECT_MAPPER.readValue(item, Organization.class)).rdd(), Encoders.bean(Organization.class)); datasource.createOrReplaceTempView("datasource"); relation.createOrReplaceTempView("relation"); @@ -107,18 +99,18 @@ public class PrepareResultCountryAssociation { String query = "SELECT source ds, country.classid country " + "FROM ( SELECT id " + - "FROM datasource " + - "WHERE (datainfo.deletedbyinference = false " + whitelisted + ") " + - getConstraintList("datasourcetype.classid = '", allowedtypes) + ") d " + + " FROM datasource " + + " WHERE (datainfo.deletedbyinference = false " + whitelisted + ") " + + getConstraintList("datasourcetype.classid = '", allowedtypes) + ") d " + "JOIN ( SELECT source, target " + - "FROM relation " + - "WHERE relclass = 'provides' " + - "AND datainfo.deletedbyinference = false ) rel " + + " FROM relation " + + " WHERE relclass = 'provides' " + + " AND datainfo.deletedbyinference = false ) rel " + "ON d.id = rel.source " + "JOIN (SELECT id, country " + - "FROM organization " + - "WHERE datainfo.deletedbyinference = false " + - "AND length(country.classid)>0) o " + + " FROM organization " + + " WHERE datainfo.deletedbyinference = false " + + " AND length(country.classid)>0) o " + "ON o.id = rel.target"; spark.sql(query) @@ -129,4 +121,6 @@ public class PrepareResultCountryAssociation { } + + } diff --git a/dhp-workflows/dhp-propagation/src/main/java/eu/dnetlib/dhp/countrypropagation/SparkCountryPropagationJob2.java b/dhp-workflows/dhp-propagation/src/main/java/eu/dnetlib/dhp/countrypropagation/SparkCountryPropagationJob2.java index 23595bada5..91b12b4445 100644 --- a/dhp-workflows/dhp-propagation/src/main/java/eu/dnetlib/dhp/countrypropagation/SparkCountryPropagationJob2.java +++ b/dhp-workflows/dhp-propagation/src/main/java/eu/dnetlib/dhp/countrypropagation/SparkCountryPropagationJob2.java @@ -4,21 +4,17 @@ import com.fasterxml.jackson.databind.ObjectMapper; import eu.dnetlib.dhp.application.ArgumentApplicationParser; import eu.dnetlib.dhp.schema.oaf.*; import org.apache.commons.io.IOUtils; -import org.apache.hadoop.fs.FileSystem; import org.apache.spark.SparkConf; import org.apache.spark.api.java.JavaPairRDD; import org.apache.spark.api.java.JavaRDD; import org.apache.spark.api.java.JavaSparkContext; +import org.apache.spark.api.java.function.MapFunction; +import org.apache.spark.broadcast.Broadcast; +import org.apache.spark.sql.*; import org.apache.spark.sql.Dataset; -import org.apache.spark.sql.Encoders; -import org.apache.spark.sql.Row; -import org.apache.spark.sql.SparkSession; import scala.Tuple2; -import java.util.ArrayList; -import java.util.Arrays; -import java.util.HashSet; -import java.util.List; +import java.util.*; import static eu.dnetlib.dhp.PropagationConstant.*; import org.slf4j.Logger; @@ -32,8 +28,25 @@ public class SparkCountryPropagationJob2 { public static void main(String[] args) throws Exception { - final ArgumentApplicationParser parser = new ArgumentApplicationParser(IOUtils.toString(SparkCountryPropagationJob2.class.getResourceAsStream("/eu/dnetlib/dhp/countrypropagation/input_countrypropagation_parameters.json"))); + String jsonConfiguration = IOUtils.toString(SparkCountryPropagationJob2.class + .getResourceAsStream("/eu/dnetlib/dhp/countrypropagation/input_countrypropagation_parameters.json")); + + final ArgumentApplicationParser parser = new ArgumentApplicationParser( + jsonConfiguration); + parser.parseArgument(args); + + String inputPath = parser.get("sourcePath"); + log.info("inputPath: {}", inputPath); + + final String outputPath = "/tmp/provision/propagation/countrytoresultfrominstitutionalrepositories"; + final String datasourcecountrypath = outputPath + "/prepared_datasource_country"; + final String resultClassName = parser.get("resultClazz"); + + final String resultType = resultClassName.substring(resultClassName.lastIndexOf(".")+1); + + Class resultClazz = (Class) Class.forName(resultClassName); + SparkConf conf = new SparkConf(); conf.set("hive.metastore.uris", parser.get("hive_metastore_uris")); final SparkSession spark = SparkSession @@ -44,242 +57,135 @@ public class SparkCountryPropagationJob2 { .enableHiveSupport() .getOrCreate(); - - final String inputPath = parser.get("sourcePath"); - final String outputPath = "/tmp/provision/propagation/countrytoresultfrominstitutionalrepositories"; - - // createOutputDirs(outputPath, FileSystem.get(spark.sparkContext().hadoopConfiguration())); - - boolean writeUpdates = TRUE.equals(parser.get("writeUpdate")); - boolean saveGraph = TRUE.equals(parser.get("saveGraph")); + final boolean writeUpdates = TRUE.equals(parser.get("writeUpdate")); + final boolean saveGraph = TRUE.equals(parser.get("saveGraph")); final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext()); + //Load parque file with preprocessed association datasource - country + Dataset datasourcecountryassoc = readAssocDatasourceCountry(spark, datasourcecountrypath); + //broadcasting the result of the preparation step + Broadcast> broadcast_datasourcecountryassoc = sc.broadcast(datasourcecountryassoc); - 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)); - - Dataset other = spark.createDataset(sc.textFile(inputPath + "/otherresearchproduct") - .map(item -> new ObjectMapper().readValue(item, OtherResearchProduct.class)).rdd(), - Encoders.bean(OtherResearchProduct.class)); - - Dataset software = spark.createDataset(sc.textFile(inputPath + "/software") - .map(item -> new ObjectMapper().readValue(item, Software.class)).rdd(), - Encoders.bean(Software.class)); - - Dataset publication = spark.createDataset(sc.textFile(inputPath + "/publication") - .map(item -> new ObjectMapper().readValue(item, Publication.class)).rdd(), - Encoders.bean(Publication.class)); - - - //todo broadcast - - - software.createOrReplaceTempView("software"); - final JavaRDD toupdateresultsoftware = propagateOnResult(spark, "software"); - - dataset.createOrReplaceTempView("dataset"); - final JavaRDD toupdateresultdataset = propagateOnResult(spark, "dataset"); - - other.createOrReplaceTempView("other"); - final JavaRDD toupdateresultother = propagateOnResult(spark, "other"); - - publication.createOrReplaceTempView("publication"); - final JavaRDD toupdateresultpublication = propagateOnResult(spark, "publication"); + Dataset potentialUpdates = getPotentialResultToUpdate(spark, inputPath, resultClazz, broadcast_datasourcecountryassoc); if(writeUpdates){ - writeUpdates(toupdateresultsoftware, toupdateresultdataset, toupdateresultother, toupdateresultpublication, outputPath); + writeUpdates(potentialUpdates.toJavaRDD(), outputPath + "/update_" + resultType); } if(saveGraph){ - createUpdateForSoftwareDataset(toupdateresultsoftware, inputPath, spark) - .map(s -> new ObjectMapper().writeValueAsString(s)) - .saveAsTextFile(outputPath + "/software"); + updateResultTable(spark, potentialUpdates, inputPath, resultClazz, outputPath + "/" + resultType); - createUpdateForDatasetDataset(toupdateresultdataset,inputPath,spark) - .map(d -> new ObjectMapper().writeValueAsString(d)) - .saveAsTextFile(outputPath + "/dataset"); - - createUpdateForOtherDataset(toupdateresultother, inputPath, spark) - .map(o -> new ObjectMapper().writeValueAsString(o)) - .saveAsTextFile(outputPath + "/otherresearchproduct"); - - createUpdateForPublicationDataset(toupdateresultpublication, inputPath, spark) - .map(p -> new ObjectMapper().writeValueAsString(p)) - .saveAsTextFile(outputPath + "/publication"); } } - private static void writeUpdates(JavaRDD software, JavaRDD dataset, JavaRDD other , JavaRDD publication, String outputPath){ - createUpdateForResultDatasetWrite(software, outputPath, "update_software"); - createUpdateForResultDatasetWrite(dataset, outputPath, "update_dataset"); - createUpdateForResultDatasetWrite(other, outputPath, "update_other"); - createUpdateForResultDatasetWrite(publication, outputPath, "update_publication"); - } + private static void updateResultTable(SparkSession spark, Dataset potentialUpdates, + String inputPath, + Class resultClazz, + String outputPath) { - private static JavaRDD createUpdateForOtherDataset(JavaRDD toupdateresult, String inputPath, SparkSession spark) { - final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext()); + Dataset result = readPathEntity(spark, inputPath, resultClazz); - return sc.textFile(inputPath + "/otherresearchproduct") - .map(item -> new ObjectMapper().readValue(item, OtherResearchProduct.class)) - .mapToPair(s -> new Tuple2<>(s.getId(), s)).leftOuterJoin(getStringResultJavaPairRDD(toupdateresult)) - .map(c -> { - OtherResearchProduct oaf = c._2()._1(); - List countryList = oaf.getCountry(); - if (c._2()._2().isPresent()) { + Dataset> result_pair = result + .map(r -> new Tuple2<>(r.getId(), r), + Encoders.tuple(Encoders.STRING(), Encoders.bean(resultClazz))); + + Dataset> potential_update_pair = potentialUpdates.map(pu -> new Tuple2<>(pu.getString(0), pu.getList(1)), + Encoders.tuple(Encoders.STRING(), Encoders.bean(List.class))); + + Dataset new_table = result_pair + .joinWith(potential_update_pair, result_pair.col("_1").equalTo(potential_update_pair.col("_1")), "left") + .map((MapFunction, Tuple2>, R>) value -> { + R r = value._1()._2(); + Optional> potentialNewCountries = Optional.ofNullable(value._2()).map(Tuple2::_2); + if (potentialNewCountries != null) { HashSet countries = new HashSet<>(); - for (Qualifier country : countryList) { + for (Qualifier country : r.getCountry()) { countries.add(country.getClassid()); } - Result r = c._2()._2().get(); - for (Country country : r.getCountry()) { - if (!countries.contains(country.getClassid())) { - countryList.add(country); + + for (Object country : potentialNewCountries.get()) { + if (!countries.contains(country)) { + r.getCountry().add(getCountry((String) country)); } } - oaf.setCountry(countryList); } - return oaf; - }); + return r; + + }, Encoders.bean(resultClazz)); + + + log.info("Saving graph table to path: {}", outputPath); + result + .toJSON() + .write() + .option("compression", "gzip") + .text(outputPath); + } + + + + private static Dataset getPotentialResultToUpdate(SparkSession spark, String inputPath, + Class resultClazz, + Broadcast> broadcast_datasourcecountryassoc) { + + Dataset result = readPathEntity(spark, inputPath, resultClazz); + result.createOrReplaceTempView("result"); + createCfHbforresult(spark); + return countryPropagationAssoc(spark, broadcast_datasourcecountryassoc); } - private static JavaRDD createUpdateForPublicationDataset(JavaRDD toupdateresult, String inputPath, SparkSession spark) { - final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext()); - return sc.textFile(inputPath + "/publication") - .map(item -> new ObjectMapper().readValue(item, Publication.class)) - .mapToPair(s -> new Tuple2<>(s.getId(), s)).leftOuterJoin(getStringResultJavaPairRDD(toupdateresult)) - .map(c -> { - Publication oaf = c._2()._1(); - List countryList = oaf.getCountry(); - if (c._2()._2().isPresent()) { - HashSet countries = new HashSet<>(); - for (Qualifier country : countryList) { - countries.add(country.getClassid()); - } - Result r = c._2()._2().get(); - for (Country country : r.getCountry()) { - if (!countries.contains(country.getClassid())) { - countryList.add(country); - } - } - oaf.setCountry(countryList); - } - return oaf; - }); - } - - private static JavaRDD createUpdateForSoftwareDataset(JavaRDD toupdateresult, String inputPath, SparkSession spark) { - final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext()); - - return sc.textFile(inputPath + "/software") - .map(item -> new ObjectMapper().readValue(item, Software.class)) - .mapToPair(s -> new Tuple2<>(s.getId(), s)).leftOuterJoin(getStringResultJavaPairRDD(toupdateresult)) - .map(c -> { - Software oaf = c._2()._1(); - List countryList = oaf.getCountry(); - if (c._2()._2().isPresent()) { - HashSet countries = new HashSet<>(); - for (Qualifier country : countryList) { - countries.add(country.getClassid()); - } - Result r = c._2()._2().get(); - for (Country country : r.getCountry()) { - if (!countries.contains(country.getClassid())) { - countryList.add(country); - } - } - oaf.setCountry(countryList); - } - return oaf; - }); - } - - private static JavaRDD createUpdateForDatasetDataset(JavaRDD toupdateresult, String inputPath, SparkSession spark) { - final JavaSparkContext sc = new JavaSparkContext(spark.sparkContext()); - - return sc.textFile(inputPath + "/dataset") - .map(item -> new ObjectMapper().readValue(item, eu.dnetlib.dhp.schema.oaf.Dataset.class)) - .mapToPair(d -> new Tuple2<>(d.getId(), d)).leftOuterJoin(getStringResultJavaPairRDD(toupdateresult)) - .map(c -> { - eu.dnetlib.dhp.schema.oaf.Dataset oaf = c._2()._1(); - List countryList = oaf.getCountry(); - if (c._2()._2().isPresent()) { - HashSet countries = new HashSet<>(); - for (Qualifier country : countryList) { - countries.add(country.getClassid()); - } - Result r = c._2()._2().get(); - for (Country country : r.getCountry()) { - if (!countries.contains(country.getClassid())) { - countryList.add(country); - } - } - oaf.setCountry(countryList); - } - return oaf; - }); - } - - private static JavaRDD propagateOnResult(SparkSession spark, String result_type) { + private static void createCfHbforresult(SparkSession spark) { String query; query = "SELECT id, inst.collectedfrom.key cf , inst.hostedby.key hb " + "FROM ( SELECT id, instance " + - "FROM " + result_type + + "FROM result " + " WHERE datainfo.deletedbyinference = false) ds " + "LATERAL VIEW EXPLODE(instance) i AS inst"; Dataset cfhb = spark.sql(query); cfhb.createOrReplaceTempView("cfhb"); - - return countryPropagationAssoc(spark, "cfhb").toJavaRDD(); - } - private static Dataset countryPropagationAssoc(SparkSession spark, String cfhbTable){ + + private static Dataset countryPropagationAssoc(SparkSession spark, + Broadcast> broadcast_datasourcecountryassoc){ + Dataset datasource_country = broadcast_datasourcecountryassoc.value(); + datasource_country.createOrReplaceTempView("datasource_country"); + String query = "SELECT id, collect_set(country) country "+ "FROM ( SELECT id, country " + "FROM rels " + - "JOIN " + cfhbTable + + "JOIN cfhb " + " ON cf = ds " + "UNION ALL " + "SELECT id , country " + "FROM rels " + - "JOIN " + cfhbTable + + "JOIN cfhb " + " ON hb = ds ) tmp " + "GROUP BY id"; return spark.sql(query); } - private static JavaPairRDD getStringResultJavaPairRDD(JavaRDD toupdateresult) { - return toupdateresult.map(c -> { - List countryList = new ArrayList<>(); - List tmp = c.getList(1); - for (String country : tmp) { - countryList.add(getCountry(country)); - } - Result r = new Result(); - r.setId(c.getString(0)); - r.setCountry(countryList); - return r; - }).mapToPair(r -> new Tuple2<>(r.getId(), r)); + private static Dataset readPathEntity(SparkSession spark, String inputEntityPath, Class resultClazz) { + + log.info("Reading Graph table from: {}", inputEntityPath); + return spark + .read() + .textFile(inputEntityPath) + .map((MapFunction) value -> OBJECT_MAPPER.readValue(value, resultClazz), Encoders.bean(resultClazz)); } - private static void createUpdateForResultDatasetWrite(JavaRDD toupdateresult, String outputPath, String type){ - toupdateresult.map(c -> { - List countryList = new ArrayList<>(); - List tmp = c.getList(1); - for (String country : tmp) { - countryList.add(getCountry(country)); - } - Result r = new Result(); - r.setId(c.getString(0)); - r.setCountry(countryList); - return r; + private static Dataset readAssocDatasourceCountry(SparkSession spark, String relationPath) { + return spark.read() + .load(relationPath) + .as(Encoders.bean(DatasourceCountry.class)); + } - }).map(r ->new ObjectMapper().writeValueAsString(r)) - .saveAsTextFile(outputPath+"/"+type); + private static void writeUpdates(JavaRDD potentialUpdates, String outputPath){ + potentialUpdates.map(u -> OBJECT_MAPPER.writeValueAsString(u)) + .saveAsTextFile(outputPath); }