graph cleaning workflow separate orcid_pending from orcid, depending on the author pid provenance

dhp-schemas/src/main/java/eu/dnetlib/dhp/schema/common/ModelConstants.java

@@ -7,6 +7,9 @@ import eu.dnetlib.dhp.schema.oaf.Qualifier;
 
 public class ModelConstants {
 
+	public static final String ORCID = "orcid";
+	public static final String ORCID_PENDING = "orcid_pending";
+
 	public static final String DNET_SUBJECT_TYPOLOGIES = "dnet:subject_classification_typologies";
 	public static final String DNET_RESULT_TYPOLOGIES = "dnet:result_typologies";
 	public static final String DNET_PUBLICATION_RESOURCE = "dnet:publication_resource";

dhp-workflows/dhp-graph-mapper/src/main/java/eu/dnetlib/dhp/oa/graph/clean/CleaningFunctions.java

@@ -189,6 +189,14 @@ public class CleaningFunctions {
 						author.setRank(i++);
 					}
 				}
+
+				final Set<String> collectedFrom = Optional
+						.ofNullable(r.getCollectedfrom())
+						.map(c -> c.stream()
+								.map(KeyValue::getKey)
+								.collect(Collectors.toCollection(HashSet::new)))
+						.orElse(new HashSet<>());
+
 				for (Author a : r.getAuthor()) {
 					if (Objects.isNull(a.getPid())) {
 						a.setPid(Lists.newArrayList());
@@ -201,13 +209,28 @@ public class CleaningFunctions {
 									.filter(p -> Objects.nonNull(p.getQualifier()))
 									.filter(p -> StringUtils.isNotBlank(p.getValue()))
 									.map(p -> {
+										// hack to distinguish orcid from orcid_pending
+										String pidProvenance = Optional
+												.ofNullable(p.getDataInfo())
+												.map(d -> Optional
+														.ofNullable(d.getProvenanceaction())
+														.map(Qualifier::getClassid)
+														.orElse(""))
+												.orElse("");
+										if (pidProvenance.equals(ModelConstants.SYSIMPORT_CROSSWALK_ENTITYREGISTRY)) {
+											p.getQualifier().setClassid(ModelConstants.ORCID);
+										} else {
+											p.getQualifier().setClassid(ModelConstants.ORCID_PENDING);
+										}
 										p.setValue(p.getValue().trim().replaceAll(ORCID_PREFIX_REGEX, ""));
 										return p;
 									})
 									.collect(
 										Collectors
 											.toMap(
-												StructuredProperty::getValue, Function.identity(), (p1, p2) -> p1,
+												p -> p.getQualifier().getClassid() + p.getValue(),
+													Function.identity(),
+													(p1, p2) -> p1,
 												LinkedHashMap::new))
 									.values()
 									.stream()

dhp-workflows/dhp-graph-mapper/src/test/resources/eu/dnetlib/dhp/oa/graph/clean/result.json

@@ -49,6 +49,28 @@
             "schemename": "dnet:pid_types"
           },
           "value": "https://orcid.org/0000-0001-9613-6639"
+        },
+        {
+          "dataInfo": {
+            "deletedbyinference": false,
+            "inferenceprovenance": "",
+            "inferred": false,
+            "invisible": false,
+            "provenanceaction": {
+              "classid": "sysimport:crosswalk:entityregistry",
+              "classname": "sysimport:crosswalk:entityregistry",
+              "schemeid": "dnet:provenanceActions",
+              "schemename": "dnet:provenanceActions"
+            },
+            "trust": "0.9"
+          },
+          "qualifier": {
+            "classid": "orcid",
+            "classname": "ORCID12",
+            "schemeid": "dnet:pid_types",
+            "schemename": "dnet:pid_types"
+          },
+          "value": "0000-0001-9613-6639"
         }
       ],
       "rank": 1,

dhp-workflows/dhp-graph-mapper/src/test/resources/eu/dnetlib/dhp/oa/graph/clean/terms.txt

@@ -1031,6 +1031,7 @@ dnet:pid_types @=@ dnet:pid_types @=@ jprn @=@ JPRN Identifier
 dnet:pid_types @=@ dnet:pid_types @=@ mag_id @=@ Microsoft Academic Graph Identifier
 dnet:pid_types @=@ dnet:pid_types @=@ oai @=@ Open Archives Initiative
 dnet:pid_types @=@ dnet:pid_types @=@ orcid @=@ Open Researcher and Contributor ID
+dnet:pid_types @=@ dnet:pid_types @=@ orcid_pending @=@ Open Researcher and Contributor ID
 dnet:pid_types @=@ dnet:pid_types @=@ PANGAEA @=@ PANGAEA
 dnet:pid_types @=@ dnet:pid_types @=@ epo_nr_epodoc @=@ Patent application number in EPODOC format
 dnet:pid_types @=@ dnet:pid_types @=@ UNKNOWN @=@ UNKNOWN