[graph cleaning] unit tests & cleanup

This commit is contained in:
Claudio Atzori 2023-04-04 12:34:51 +02:00
parent b502f86523
commit 2a6ba29b64
3 changed files with 75 additions and 57 deletions

View File

@ -16,6 +16,8 @@ import org.apache.commons.lang3.StringUtils;
import org.apache.spark.api.java.function.MapFunction;
import org.apache.spark.sql.Encoders;
import com.fasterxml.jackson.core.JsonProcessingException;
import com.fasterxml.jackson.databind.ObjectMapper;
import com.github.sisyphsu.dateparser.DateParserUtils;
import com.google.common.collect.Lists;
import com.google.common.collect.Sets;
@ -41,7 +43,23 @@ public class GraphCleaningFunctions extends CleaningFunctions {
public static <T extends Oaf> T cleanContext(T value, String contextId, String verifyParam) {
if (ModelSupport.isSubClass(value, Result.class)) {
final Result res = (Result) value;
if (res
if (shouldCleanContext(res, verifyParam)) {
res
.setContext(
res
.getContext()
.stream()
.filter(c -> !StringUtils.startsWith(c.getId().toLowerCase(), contextId))
.collect(Collectors.toList()));
}
return (T) res;
} else {
return value;
}
}
private static boolean shouldCleanContext(Result res, String verifyParam) {
boolean titleMatch = res
.getTitle()
.stream()
.filter(
@ -49,22 +67,9 @@ public class GraphCleaningFunctions extends CleaningFunctions {
.getQualifier()
.getClassid()
.equalsIgnoreCase(ModelConstants.MAIN_TITLE_QUALIFIER.getClassid()))
.noneMatch(t -> t.getValue().toLowerCase().startsWith(verifyParam.toLowerCase()))) {
return (T) res;
}
res
.setContext(
res
.getContext()
.stream()
.filter(
c -> !c.getId().split("::")[0]
.equalsIgnoreCase(contextId))
.collect(Collectors.toList()));
return (T) res;
} else {
return value;
}
.anyMatch(t -> t.getValue().toLowerCase().startsWith(verifyParam.toLowerCase()));
return titleMatch && Objects.nonNull(res.getContext());
}
public static <T extends Oaf> T cleanCountry(T value, String[] verifyParam, Set<String> hostedBy,

View File

@ -149,15 +149,6 @@ public class CleanGraphSparkJob {
.map((MapFunction<T, T>) value -> GraphCleaningFunctions.cleanup(value, vocs), Encoders.bean(clazz))
.filter((FilterFunction<T>) GraphCleaningFunctions::filter);
if (Boolean.FALSE.equals(deepClean)) {
cleaned_basic
.write()
.mode(SaveMode.Overwrite)
.option("compression", "gzip")
.json(outputPath);
} else if (Boolean.TRUE.equals(ModelSupport.isSubClass(clazz, Result.class))) {
// read the master-duplicate tuples
Dataset<MasterDuplicate> md = spark
.read()
@ -171,11 +162,15 @@ public class CleanGraphSparkJob {
.map(as(clazz), Encoders.bean(clazz))
.flatMap(flattenCfHbFn(), Encoders.bean(IdCfHbMapping.class));
// set the EMPTY master ID/NAME
Dataset<IdCfHbMapping> resolvedDs = resolved
.joinWith(md, resolved.col("cfhb").equalTo(md.col("duplicateId")))
.map(asIdCfHbMapping(), Encoders.bean(IdCfHbMapping.class))
.filter((FilterFunction<IdCfHbMapping>) m -> Objects.nonNull(m.getMasterId()));
if (Boolean.FALSE.equals(deepClean)) {
if (Boolean.TRUE.equals(ModelSupport.isSubClass(clazz, Result.class))) {
save(fixCFHB(clazz, cleaned_basic, md, resolved), outputPath);
} else {
save(cleaned_basic, outputPath);
}
} else if (Boolean.TRUE.equals(ModelSupport.isSubClass(clazz, Result.class))) {
// load the hostedby mapping
Set<String> hostedBy = Sets
@ -186,7 +181,7 @@ public class CleanGraphSparkJob {
.collectAsList());
// perform the deep cleaning steps
final Dataset<T> cleaned_deep = cleaned_basic
final Dataset<T> cleaned_deep = fixCFHB(clazz, cleaned_basic, md, resolved)
.map(
(MapFunction<T, T>) value -> GraphCleaningFunctions.cleanContext(value, contextId, verifyParam),
Encoders.bean(clazz))
@ -195,17 +190,32 @@ public class CleanGraphSparkJob {
.cleanCountry(value, verifyCountryParam, hostedBy, collectedfrom, country),
Encoders.bean(clazz));
// Join the results with the resolved CF|HB mapping, apply the mapping and save it
cleaned_deep
.joinWith(resolvedDs, cleaned_deep.col("id").equalTo(resolvedDs.col("resultId")), "left")
.groupByKey(
(MapFunction<Tuple2<T, IdCfHbMapping>, String>) t -> ((Result) t._1()).getId(), Encoders.STRING())
.mapGroups(getMapGroupsFunction(), Encoders.bean(clazz))
save(cleaned_deep, outputPath);
}
}
private static <T extends Oaf> void save(final Dataset<T> dataset, final String outputPath) {
dataset
.write()
.mode(SaveMode.Overwrite)
.option("compression", "gzip")
.json(outputPath);
}
private static <T extends Oaf> Dataset<T> fixCFHB(Class<T> clazz, Dataset<T> results, Dataset<MasterDuplicate> md,
Dataset<IdCfHbMapping> resolved) {
// set the EMPTY master ID/NAME
Dataset<IdCfHbMapping> resolvedDs = resolved
.joinWith(md, resolved.col("cfhb").equalTo(md.col("duplicateId")))
.map(asIdCfHbMapping(), Encoders.bean(IdCfHbMapping.class))
.filter((FilterFunction<IdCfHbMapping>) m -> Objects.nonNull(m.getMasterId()));
return results
.joinWith(resolvedDs, results.col("id").equalTo(resolvedDs.col("resultId")), "left")
.groupByKey(
(MapFunction<Tuple2<T, IdCfHbMapping>, String>) t -> ((Result) t._1()).getId(), Encoders.STRING())
.mapGroups(getMapGroupsFunction(), Encoders.bean(clazz));
}
private static <T extends Oaf> Dataset<T> readTableFromPath(

View File

@ -165,7 +165,8 @@ public class CleanGraphSparkJobTest {
"--outputPath", graphOutputPath + "/relation",
"--isLookupUrl", "lookupurl",
"--graphTableClassName", Relation.class.getCanonicalName(),
"--deepClean", "false"
"--deepClean", "false",
"--masterDuplicatePath", dsMasterDuplicatePath,
})).run(false, isLookUpService);
spark
@ -262,7 +263,8 @@ public class CleanGraphSparkJobTest {
"--outputPath", graphOutputPath + "/publication",
"--isLookupUrl", "lookupurl",
"--graphTableClassName", Publication.class.getCanonicalName(),
"--deepClean", "false"
"--deepClean", "false",
"--masterDuplicatePath", dsMasterDuplicatePath,
})).run(false, isLookUpService);
Publication p = read(spark, graphOutputPath + "/publication", Publication.class)
@ -413,7 +415,8 @@ public class CleanGraphSparkJobTest {
"--outputPath", graphOutputPath + "/publication",
"--isLookupUrl", "lookupurl",
"--graphTableClassName", Publication.class.getCanonicalName(),
"--deepClean", "false"
"--deepClean", "false",
"--masterDuplicatePath", dsMasterDuplicatePath,
})).run(false, isLookUpService);
Dataset<Publication> p = read(spark, graphOutputPath + "/publication", Publication.class)