[DOIBoost Mapping] Fix to avoid to produce results with null as identifier (probably due to the filtering function in the factory for the creation of the id)

dhp-workflows/dhp-doiboost/src/main/java/eu/dnetlib/doiboost/mag/SparkProcessMAG.scala

@@ -164,12 +164,18 @@ object SparkProcessMAG {
       .write.mode(SaveMode.Overwrite)
       .save(s"$workingPath/mag_publication")
 
-
-    val s:RDD[Publication] = spark.read.load(s"$workingPath/mag_publication").as[Publication]
-      .map(p=>Tuple2(p.getId, p)).rdd.reduceByKey((a:Publication, b:Publication) => ConversionUtil.mergePublication(a,b))
-    .map(_._2)
-
-    spark.createDataset(s).as[Publication].write.mode(SaveMode.Overwrite).save(s"$targetPath/magPublication")
+    spark.read.load(s"$workingPath/mag_publication").as[Publication]
+      .filter(p => p.getId == null)
+      .groupByKey(p => p.getId)
+      .reduceGroups((a:Publication, b:Publication) => ConversionUtil.mergePublication(a,b))
+      .map(_._2)
+      .write.mode(SaveMode.Overwrite).save(s"$targetPath/magPublication")
+
+//    val s:RDD[Publication] = spark.read.load(s"$workingPath/mag_publication").as[Publication]
+//      .map(p=>Tuple2(p.getId, p)).rdd.reduceByKey((a:Publication, b:Publication) => ConversionUtil.mergePublication(a,b))
+//    .map(_._2)
+//
+//    spark.createDataset(s).as[Publication].write.mode(SaveMode.Overwrite).save(s"$targetPath/magPublication")
 
   }
 }