forked from D-Net/dnet-hadoop
[aggregator graph] save invalid records aside for further inspection
This commit is contained in:
parent
b99a011345
commit
a0919ed495
|
@ -107,7 +107,7 @@ public abstract class AbstractMdRecordToOafMapper {
|
||||||
this.forceOriginalId = false;
|
this.forceOriginalId = false;
|
||||||
}
|
}
|
||||||
|
|
||||||
public List<Oaf> processMdRecord(final String xml) throws DocumentException {
|
public List<Oaf> processMdRecord(final String xml) {
|
||||||
|
|
||||||
DocumentFactory.getInstance().setXPathNamespaceURIs(nsContext);
|
DocumentFactory.getInstance().setXPathNamespaceURIs(nsContext);
|
||||||
try {
|
try {
|
||||||
|
|
|
@ -16,6 +16,9 @@ import org.apache.hadoop.io.compress.GzipCodec;
|
||||||
import org.apache.spark.SparkConf;
|
import org.apache.spark.SparkConf;
|
||||||
import org.apache.spark.api.java.JavaRDD;
|
import org.apache.spark.api.java.JavaRDD;
|
||||||
import org.apache.spark.api.java.JavaSparkContext;
|
import org.apache.spark.api.java.JavaSparkContext;
|
||||||
|
import org.apache.spark.rdd.RDD;
|
||||||
|
import org.apache.spark.sql.Encoders;
|
||||||
|
import org.apache.spark.sql.SaveMode;
|
||||||
import org.apache.spark.sql.SparkSession;
|
import org.apache.spark.sql.SparkSession;
|
||||||
import org.dom4j.DocumentException;
|
import org.dom4j.DocumentException;
|
||||||
import org.slf4j.Logger;
|
import org.slf4j.Logger;
|
||||||
|
@ -76,6 +79,9 @@ public class GenerateEntitiesApplication {
|
||||||
final String targetPath = parser.get("targetPath");
|
final String targetPath = parser.get("targetPath");
|
||||||
log.info("targetPath: {}", targetPath);
|
log.info("targetPath: {}", targetPath);
|
||||||
|
|
||||||
|
final String invalidPath = parser.get("invalidPath");
|
||||||
|
log.info("invalidPath: {}", invalidPath);
|
||||||
|
|
||||||
final String isLookupUrl = parser.get("isLookupUrl");
|
final String isLookupUrl = parser.get("isLookupUrl");
|
||||||
log.info("isLookupUrl: {}", isLookupUrl);
|
log.info("isLookupUrl: {}", isLookupUrl);
|
||||||
|
|
||||||
|
@ -97,7 +103,8 @@ public class GenerateEntitiesApplication {
|
||||||
final SparkConf conf = new SparkConf();
|
final SparkConf conf = new SparkConf();
|
||||||
runWithSparkSession(conf, isSparkSessionManaged, spark -> {
|
runWithSparkSession(conf, isSparkSessionManaged, spark -> {
|
||||||
HdfsSupport.remove(targetPath, spark.sparkContext().hadoopConfiguration());
|
HdfsSupport.remove(targetPath, spark.sparkContext().hadoopConfiguration());
|
||||||
generateEntities(spark, vocs, sourcePaths, targetPath, shouldHashId, mode);
|
HdfsSupport.remove(invalidPath, spark.sparkContext().hadoopConfiguration());
|
||||||
|
generateEntities(spark, vocs, sourcePaths, targetPath, invalidPath, shouldHashId, mode);
|
||||||
});
|
});
|
||||||
}
|
}
|
||||||
|
|
||||||
|
@ -106,6 +113,7 @@ public class GenerateEntitiesApplication {
|
||||||
final VocabularyGroup vocs,
|
final VocabularyGroup vocs,
|
||||||
final String sourcePaths,
|
final String sourcePaths,
|
||||||
final String targetPath,
|
final String targetPath,
|
||||||
|
final String invalidPath,
|
||||||
final boolean shouldHashId,
|
final boolean shouldHashId,
|
||||||
final Mode mode) {
|
final Mode mode) {
|
||||||
|
|
||||||
|
@ -121,6 +129,19 @@ public class GenerateEntitiesApplication {
|
||||||
JavaRDD<Oaf> inputRdd = sc.emptyRDD();
|
JavaRDD<Oaf> inputRdd = sc.emptyRDD();
|
||||||
|
|
||||||
for (final String sp : existingSourcePaths) {
|
for (final String sp : existingSourcePaths) {
|
||||||
|
RDD<String> invalidRecords = sc
|
||||||
|
.sequenceFile(sp, Text.class, Text.class)
|
||||||
|
.map(k -> new Tuple2<>(k._1().toString(), k._2().toString()))
|
||||||
|
.map(k -> tryApplyMapping(k._1(), k._2(), shouldHashId, vocs))
|
||||||
|
.filter(Objects::nonNull)
|
||||||
|
.rdd();
|
||||||
|
spark
|
||||||
|
.createDataset(invalidRecords, Encoders.STRING())
|
||||||
|
.write()
|
||||||
|
.mode(SaveMode.Append)
|
||||||
|
.option("compression", "gzip")
|
||||||
|
.text(invalidPath);
|
||||||
|
|
||||||
inputRdd = inputRdd
|
inputRdd = inputRdd
|
||||||
.union(
|
.union(
|
||||||
sc
|
sc
|
||||||
|
@ -159,7 +180,7 @@ public class GenerateEntitiesApplication {
|
||||||
final String id,
|
final String id,
|
||||||
final String s,
|
final String s,
|
||||||
final boolean shouldHashId,
|
final boolean shouldHashId,
|
||||||
final VocabularyGroup vocs) throws DocumentException {
|
final VocabularyGroup vocs) {
|
||||||
final String type = StringUtils.substringAfter(id, ":");
|
final String type = StringUtils.substringAfter(id, ":");
|
||||||
|
|
||||||
switch (type.toLowerCase()) {
|
switch (type.toLowerCase()) {
|
||||||
|
@ -196,6 +217,18 @@ public class GenerateEntitiesApplication {
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
private static String tryApplyMapping(
|
||||||
|
final String id,
|
||||||
|
final String s,
|
||||||
|
final boolean shouldHashId,
|
||||||
|
final VocabularyGroup vocs) {
|
||||||
|
|
||||||
|
if (convertToListOaf(id, s, shouldHashId, vocs).isEmpty()) {
|
||||||
|
return s;
|
||||||
|
}
|
||||||
|
return null;
|
||||||
|
}
|
||||||
|
|
||||||
private static Oaf convertFromJson(final String s, final Class<? extends Oaf> clazz) {
|
private static Oaf convertFromJson(final String s, final Class<? extends Oaf> clazz) {
|
||||||
try {
|
try {
|
||||||
return OBJECT_MAPPER.readValue(s, clazz);
|
return OBJECT_MAPPER.readValue(s, clazz);
|
||||||
|
|
|
@ -17,6 +17,12 @@
|
||||||
"paramDescription": "the path of the target file",
|
"paramDescription": "the path of the target file",
|
||||||
"paramRequired": true
|
"paramRequired": true
|
||||||
},
|
},
|
||||||
|
{
|
||||||
|
"paramName": "i",
|
||||||
|
"paramLongName": "invalidPath",
|
||||||
|
"paramDescription": "the path of the invalid records file",
|
||||||
|
"paramRequired": false
|
||||||
|
},
|
||||||
{
|
{
|
||||||
"paramName": "isu",
|
"paramName": "isu",
|
||||||
"paramLongName": "isLookupUrl",
|
"paramLongName": "isLookupUrl",
|
||||||
|
|
|
@ -468,6 +468,7 @@
|
||||||
</spark-opts>
|
</spark-opts>
|
||||||
<arg>--sourcePaths</arg><arg>${contentPath}/db_claims,${contentPath}/oaf_claims,${contentPath}/odf_claims</arg>
|
<arg>--sourcePaths</arg><arg>${contentPath}/db_claims,${contentPath}/oaf_claims,${contentPath}/odf_claims</arg>
|
||||||
<arg>--targetPath</arg><arg>${workingDir}/entities_claim</arg>
|
<arg>--targetPath</arg><arg>${workingDir}/entities_claim</arg>
|
||||||
|
<arg>--invalidPath</arg><arg>${workingDir}/invalid_records_claim</arg>
|
||||||
<arg>--isLookupUrl</arg><arg>${isLookupUrl}</arg>
|
<arg>--isLookupUrl</arg><arg>${isLookupUrl}</arg>
|
||||||
<arg>--shouldHashId</arg><arg>${shouldHashId}</arg>
|
<arg>--shouldHashId</arg><arg>${shouldHashId}</arg>
|
||||||
<arg>--mode</arg><arg>claim</arg>
|
<arg>--mode</arg><arg>claim</arg>
|
||||||
|
@ -517,6 +518,7 @@
|
||||||
</spark-opts>
|
</spark-opts>
|
||||||
<arg>--sourcePaths</arg><arg>${contentPath}/db_openaire,${contentPath}/db_openorgs,${contentPath}/oaf_records,${contentPath}/odf_records,${contentPath}/oaf_records_hdfs,${contentPath}/odf_records_hdfs,${contentPath}/oaf_records_invisible</arg>
|
<arg>--sourcePaths</arg><arg>${contentPath}/db_openaire,${contentPath}/db_openorgs,${contentPath}/oaf_records,${contentPath}/odf_records,${contentPath}/oaf_records_hdfs,${contentPath}/odf_records_hdfs,${contentPath}/oaf_records_invisible</arg>
|
||||||
<arg>--targetPath</arg><arg>${workingDir}/entities</arg>
|
<arg>--targetPath</arg><arg>${workingDir}/entities</arg>
|
||||||
|
<arg>--invalidPath</arg><arg>${workingDir}/invalid_records</arg>
|
||||||
<arg>--isLookupUrl</arg><arg>${isLookupUrl}</arg>
|
<arg>--isLookupUrl</arg><arg>${isLookupUrl}</arg>
|
||||||
<arg>--shouldHashId</arg><arg>${shouldHashId}</arg>
|
<arg>--shouldHashId</arg><arg>${shouldHashId}</arg>
|
||||||
</spark>
|
</spark>
|
||||||
|
|
|
@ -12,7 +12,6 @@ import java.util.Objects;
|
||||||
import java.util.Optional;
|
import java.util.Optional;
|
||||||
import java.util.stream.Collectors;
|
import java.util.stream.Collectors;
|
||||||
|
|
||||||
import kotlin.jvm.Throws;
|
|
||||||
import org.apache.commons.io.IOUtils;
|
import org.apache.commons.io.IOUtils;
|
||||||
import org.apache.commons.lang3.StringUtils;
|
import org.apache.commons.lang3.StringUtils;
|
||||||
import org.dom4j.DocumentException;
|
import org.dom4j.DocumentException;
|
||||||
|
@ -22,7 +21,6 @@ import org.junit.jupiter.api.extension.ExtendWith;
|
||||||
import org.mockito.Mock;
|
import org.mockito.Mock;
|
||||||
import org.mockito.junit.jupiter.MockitoExtension;
|
import org.mockito.junit.jupiter.MockitoExtension;
|
||||||
|
|
||||||
import com.fasterxml.jackson.core.JsonProcessingException;
|
|
||||||
import com.fasterxml.jackson.databind.ObjectMapper;
|
import com.fasterxml.jackson.databind.ObjectMapper;
|
||||||
|
|
||||||
import eu.dnetlib.dhp.common.vocabulary.VocabularyGroup;
|
import eu.dnetlib.dhp.common.vocabulary.VocabularyGroup;
|
||||||
|
|
|
@ -18,7 +18,6 @@ import java.util.Objects;
|
||||||
import java.util.Optional;
|
import java.util.Optional;
|
||||||
import java.util.stream.Collectors;
|
import java.util.stream.Collectors;
|
||||||
|
|
||||||
import eu.dnetlib.dhp.schema.common.ModelConstants;
|
|
||||||
import org.apache.commons.io.IOUtils;
|
import org.apache.commons.io.IOUtils;
|
||||||
import org.apache.commons.lang3.StringUtils;
|
import org.apache.commons.lang3.StringUtils;
|
||||||
import org.junit.jupiter.api.BeforeEach;
|
import org.junit.jupiter.api.BeforeEach;
|
||||||
|
@ -32,6 +31,7 @@ import com.fasterxml.jackson.core.type.TypeReference;
|
||||||
import com.fasterxml.jackson.databind.ObjectMapper;
|
import com.fasterxml.jackson.databind.ObjectMapper;
|
||||||
|
|
||||||
import eu.dnetlib.dhp.common.vocabulary.VocabularyGroup;
|
import eu.dnetlib.dhp.common.vocabulary.VocabularyGroup;
|
||||||
|
import eu.dnetlib.dhp.schema.common.ModelConstants;
|
||||||
import eu.dnetlib.dhp.schema.oaf.*;
|
import eu.dnetlib.dhp.schema.oaf.*;
|
||||||
import eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils;
|
import eu.dnetlib.dhp.schema.oaf.utils.OafMapperUtils;
|
||||||
|
|
||||||
|
|
Loading…
Reference in New Issue