1 |
26600
|
sandro.lab
|
package eu.dnetlib.data.transform.xml;
|
2 |
|
|
|
3 |
52411
|
claudio.at
|
import java.util.HashMap;
|
4 |
26600
|
sandro.lab
|
import java.util.List;
|
5 |
|
|
|
6 |
52231
|
claudio.at
|
import com.google.common.collect.Iterables;
|
7 |
26600
|
sandro.lab
|
import com.google.common.collect.Lists;
|
8 |
|
|
import com.google.protobuf.Descriptors.Descriptor;
|
9 |
|
|
import eu.dnetlib.data.mapreduce.util.OafRowKeyDecoder;
|
10 |
49028
|
claudio.at
|
import eu.dnetlib.data.proto.FieldTypeProtos.Author;
|
11 |
42532
|
alessia.ba
|
import eu.dnetlib.data.proto.FieldTypeProtos.KeyValue;
|
12 |
28092
|
claudio.at
|
import eu.dnetlib.data.proto.FieldTypeProtos.StructuredProperty;
|
13 |
26600
|
sandro.lab
|
import eu.dnetlib.data.proto.OafProtos.Oaf;
|
14 |
|
|
import eu.dnetlib.data.proto.OafProtos.OafEntity;
|
15 |
|
|
import eu.dnetlib.data.proto.ResultProtos.Result;
|
16 |
|
|
import eu.dnetlib.data.proto.ResultProtos.Result.Context;
|
17 |
|
|
import eu.dnetlib.data.proto.ResultProtos.Result.ExternalReference;
|
18 |
|
|
import eu.dnetlib.data.proto.ResultProtos.Result.Instance;
|
19 |
|
|
import eu.dnetlib.data.proto.TypeProtos.Type;
|
20 |
52169
|
claudio.at
|
import org.apache.commons.lang3.StringUtils;
|
21 |
40014
|
claudio.at
|
import org.w3c.dom.NodeList;
|
22 |
26600
|
sandro.lab
|
|
23 |
40198
|
claudio.at
|
public class OafToHbaseXsltFunctions extends CommonDNetXsltFunctions {
|
24 |
26600
|
sandro.lab
|
|
25 |
40198
|
claudio.at
|
public static String oafResult(
|
26 |
|
|
final String resultId,
|
27 |
49480
|
claudio.at
|
final boolean invisible,
|
28 |
40198
|
claudio.at
|
final String provenance,
|
29 |
30969
|
claudio.at
|
final String trust,
|
30 |
38972
|
claudio.at
|
final NodeList about,
|
31 |
26600
|
sandro.lab
|
final String originalId,
|
32 |
|
|
final String dateOfCollection,
|
33 |
40313
|
claudio.at
|
final String dateOfTransformation,
|
34 |
52169
|
claudio.at
|
final NodeList metadata) {
|
35 |
|
|
|
36 |
|
|
ValueMap values = null;
|
37 |
26600
|
sandro.lab
|
try {
|
38 |
|
|
final String entityId = OafRowKeyDecoder.decode(resultId).getKey();
|
39 |
52169
|
claudio.at
|
values = ValueMap.parseNodeList(metadata);
|
40 |
42532
|
alessia.ba
|
final Descriptor mDesc = Result.Metadata.getDescriptor();
|
41 |
26600
|
sandro.lab
|
|
42 |
52169
|
claudio.at
|
final List<KeyValue> collectedFrom = getKeyValues(values, "collectedfrom", Type.datasource);
|
43 |
|
|
final List<KeyValue> hostedBy = getKeyValues(values, "hostedby", Type.datasource);
|
44 |
|
|
|
45 |
|
|
final Result.Metadata.Builder metadataBuilder = buildMetadata(values, mDesc);
|
46 |
|
|
final Result.Builder result = buildResult(metadataBuilder, values, collectedFrom, hostedBy);
|
47 |
|
|
final OafEntity.Builder entity = buildOafEntity(result, entityId, metadata, collectedFrom, originalId);
|
48 |
42532
|
alessia.ba
|
entity.setDateofcollection(dateOfCollection)
|
49 |
|
|
.setDateoftransformation(dateOfTransformation).setOaiprovenance(getOAIProvenance(about));
|
50 |
26600
|
sandro.lab
|
|
51 |
49480
|
claudio.at
|
final Oaf oaf = getOaf(entity, getDataInfo(invisible, about, provenance, trust, false, false));
|
52 |
42532
|
alessia.ba
|
return base64(oaf.toByteArray());
|
53 |
|
|
} catch (final Throwable e) {
|
54 |
52169
|
claudio.at
|
handleException(e, resultId, values);
|
55 |
42532
|
alessia.ba
|
}
|
56 |
|
|
return null;
|
57 |
|
|
}
|
58 |
|
|
|
59 |
|
|
public static String oafResultUpdate(final String resultId,
|
60 |
|
|
final String provenance,
|
61 |
|
|
final String trust,
|
62 |
52169
|
claudio.at
|
final NodeList nodelist) {
|
63 |
|
|
ValueMap values = null;
|
64 |
42532
|
alessia.ba
|
try {
|
65 |
|
|
final String entityId = OafRowKeyDecoder.decode(resultId).getKey();
|
66 |
52169
|
claudio.at
|
values = ValueMap.parseNodeList(nodelist);
|
67 |
|
|
final List<KeyValue> hostedBy = getKeyValues(values, "hostedby", Type.datasource);
|
68 |
|
|
|
69 |
26600
|
sandro.lab
|
final Descriptor mDesc = Result.Metadata.getDescriptor();
|
70 |
|
|
|
71 |
42532
|
alessia.ba
|
final Result.Metadata.Builder metadata = buildMetadata(values, mDesc);
|
72 |
48679
|
claudio.at
|
final Result.Builder result = buildResult(metadata, values, null, hostedBy);
|
73 |
35824
|
claudio.at
|
|
74 |
42532
|
alessia.ba
|
final OafEntity.Builder entity = buildOafEntity(result, entityId, nodelist, null, null);
|
75 |
|
|
final Oaf oaf = getOaf(entity, null);
|
76 |
|
|
return base64(oaf.toByteArray());
|
77 |
|
|
} catch (final Throwable e) {
|
78 |
52169
|
claudio.at
|
handleException(e, resultId, values);
|
79 |
42532
|
alessia.ba
|
}
|
80 |
|
|
return null;
|
81 |
|
|
}
|
82 |
35824
|
claudio.at
|
|
83 |
52169
|
claudio.at
|
private static OafEntity.Builder buildOafEntity(
|
84 |
|
|
final Result.Builder result,
|
85 |
42532
|
alessia.ba
|
final String entityId,
|
86 |
|
|
final NodeList nodelist,
|
87 |
52169
|
claudio.at
|
final List<KeyValue> collectedFrom,
|
88 |
|
|
final String originalId) {
|
89 |
35824
|
claudio.at
|
|
90 |
42532
|
alessia.ba
|
final List<StructuredProperty> pids = Lists.newArrayList();
|
91 |
|
|
pids.addAll(parsePids(nodelist));
|
92 |
35824
|
claudio.at
|
|
93 |
42532
|
alessia.ba
|
final OafEntity.Builder entity =
|
94 |
|
|
getEntity(Type.result, entityId, collectedFrom, StringUtils.isBlank(originalId) ? null : Lists.newArrayList(originalId), null, null, pids)
|
95 |
|
|
.setResult(result);
|
96 |
|
|
return entity;
|
97 |
|
|
}
|
98 |
|
|
|
99 |
|
|
private static Result.Metadata.Builder buildMetadata(final ValueMap values, final Descriptor mDesc) {
|
100 |
|
|
final Result.Metadata.Builder metadata = Result.Metadata.newBuilder();
|
101 |
43986
|
claudio.at
|
|
102 |
49028
|
claudio.at
|
if (values.get("creator") != null) {
|
103 |
|
|
for (final Element e : values.get("creator")) {
|
104 |
|
|
|
105 |
|
|
final Author.Builder author = Author.newBuilder();
|
106 |
|
|
|
107 |
|
|
final String fullname = e.getText();
|
108 |
|
|
author.setFullname(fullname);
|
109 |
|
|
author.setRank(Integer.valueOf(e.getAttributeValue(ValueMap.IDX_ATTRIBUTE)));
|
110 |
|
|
|
111 |
52169
|
claudio.at
|
final String nameIdentifier = e.getAttributeValue("nameIdentifier");
|
112 |
|
|
final String nameIdentifierScheme = e.getAttributeValue("nameIdentifierScheme");
|
113 |
|
|
|
114 |
|
|
if (StringUtils.isNotBlank(nameIdentifier) && StringUtils.isNotBlank(nameIdentifierScheme)) {
|
115 |
|
|
author.addPid(getKV(nameIdentifierScheme, nameIdentifier));
|
116 |
|
|
}
|
117 |
|
|
|
118 |
49028
|
claudio.at
|
final eu.dnetlib.pace.model.Person p = new eu.dnetlib.pace.model.Person(fullname, false);
|
119 |
|
|
if (p.isAccurate()) {
|
120 |
|
|
author.setName(p.getNormalisedFirstName());
|
121 |
|
|
author.setSurname(p.getNormalisedSurname());
|
122 |
|
|
}
|
123 |
|
|
metadata.addAuthor(author);
|
124 |
|
|
}
|
125 |
|
|
}
|
126 |
|
|
|
127 |
44552
|
claudio.at
|
addStructuredProps(metadata, mDesc.findFieldByName("subject"), values.get("subject"), "keyword", "dnet:subject_classification_typologies");
|
128 |
43986
|
claudio.at
|
addStructuredProps(metadata, mDesc.findFieldByName("title"), values.get("title"), "main title", "dnet:dataCite_title");
|
129 |
42532
|
alessia.ba
|
for (final String fieldname : Lists.newArrayList("description", "source", "contributor")) {
|
130 |
|
|
if (values.get(fieldname) != null) {
|
131 |
|
|
for (final String s : values.get(fieldname).listValues()) {
|
132 |
|
|
addField(metadata, mDesc.findFieldByName(fieldname), s);
|
133 |
35824
|
claudio.at
|
}
|
134 |
|
|
}
|
135 |
42532
|
alessia.ba
|
}
|
136 |
|
|
addField(metadata, mDesc.findFieldByName("language"), setQualifier(getDefaultQualifier("dnet:languages"), values.get("language").listValues()));
|
137 |
|
|
addField(metadata, mDesc.findFieldByName("dateofacceptance"), values.get("dateaccepted").listValues());
|
138 |
|
|
addField(metadata, mDesc.findFieldByName("publisher"), values.get("publisher").listValues());
|
139 |
|
|
addField(metadata, mDesc.findFieldByName("embargoenddate"), values.get("embargoenddate").listValues());
|
140 |
|
|
addField(metadata, mDesc.findFieldByName("storagedate"), values.get("storagedate").listValues());
|
141 |
35824
|
claudio.at
|
|
142 |
52411
|
claudio.at
|
String resulttype = getResultType(values);
|
143 |
47480
|
claudio.at
|
addField(metadata, mDesc.findFieldByName("resulttype"), getSimpleQualifier(resulttype, "dnet:result_typologies"));
|
144 |
|
|
|
145 |
42532
|
alessia.ba
|
addField(metadata, mDesc.findFieldByName("fulltext"), values.get("fulltext").listValues());
|
146 |
|
|
addField(metadata, mDesc.findFieldByName("format"), values.get("format").listValues());
|
147 |
|
|
if (values.get("concept") != null) {
|
148 |
|
|
for (final Element e : values.get("concept")) {
|
149 |
|
|
final String id = e.getAttributes().get("id");
|
150 |
52945
|
claudio.at
|
if (StringUtils.isNotBlank(id)) {
|
151 |
|
|
metadata.addContext(Context.newBuilder().setId(id));
|
152 |
|
|
}
|
153 |
26600
|
sandro.lab
|
}
|
154 |
42532
|
alessia.ba
|
}
|
155 |
|
|
if (values.get("journal") != null) {
|
156 |
|
|
for (final Element e : values.get("journal")) {
|
157 |
54977
|
alessia.ba
|
addJournal(metadata, e);
|
158 |
26600
|
sandro.lab
|
}
|
159 |
42532
|
alessia.ba
|
}
|
160 |
|
|
return metadata;
|
161 |
|
|
}
|
162 |
26600
|
sandro.lab
|
|
163 |
52411
|
claudio.at
|
private static String getResultType(final ValueMap values) {
|
164 |
|
|
|
165 |
|
|
final Element cobjcategory = values.get("cobjcategory").stream()
|
166 |
|
|
.map(e -> StringUtils.isNotBlank(e.getText()) ? e : new Element("0000", e.getAttributes()))
|
167 |
|
|
.findFirst()
|
168 |
|
|
.orElse(new Element("0000", new HashMap<>()));
|
169 |
|
|
|
170 |
|
|
final String resulttype = cobjcategory.getAttributeValue("type");
|
171 |
|
|
if (StringUtils.isNotBlank(resulttype)) {
|
172 |
|
|
return resulttype;
|
173 |
|
|
}
|
174 |
|
|
|
175 |
52525
|
claudio.at
|
return getDefaultResulttype(cobjcategory);
|
176 |
52411
|
claudio.at
|
}
|
177 |
|
|
|
178 |
42532
|
alessia.ba
|
private static Result.Builder buildResult(final Result.Metadata.Builder metadata,
|
179 |
|
|
final ValueMap values,
|
180 |
52169
|
claudio.at
|
final List<KeyValue> collectedFrom,
|
181 |
|
|
final List<KeyValue> hostedBy) {
|
182 |
42532
|
alessia.ba
|
final Result.Builder result = Result.newBuilder();
|
183 |
49028
|
claudio.at
|
|
184 |
52169
|
claudio.at
|
final Instance.Builder instance = Instance.newBuilder();
|
185 |
26600
|
sandro.lab
|
|
186 |
49095
|
claudio.at
|
addField(instance, Instance.getDescriptor().findFieldByName("license"), values.get("license").listValues());
|
187 |
|
|
|
188 |
|
|
addField(instance, Instance.getDescriptor().findFieldByName("accessright"),
|
189 |
42532
|
alessia.ba
|
setQualifier(getDefaultQualifier("dnet:access_modes"), values.get("accessrights").listValues()));
|
190 |
49095
|
claudio.at
|
|
191 |
42532
|
alessia.ba
|
addField(instance, Instance.getDescriptor().findFieldByName("instancetype"),
|
192 |
|
|
setQualifier(getDefaultQualifier("dnet:publication_resource"), values.get("cobjcategory").listValues()));
|
193 |
26600
|
sandro.lab
|
|
194 |
52169
|
claudio.at
|
addField(instance, Instance.getDescriptor().findFieldByName("hostedby"), hostedBy);
|
195 |
48679
|
claudio.at
|
addField(instance, Instance.getDescriptor().findFieldByName("collectedfrom"), collectedFrom);
|
196 |
|
|
addField(instance, Instance.getDescriptor().findFieldByName("dateofacceptance"), values.get("dateaccepted").listValues());
|
197 |
|
|
|
198 |
42532
|
alessia.ba
|
if (values.get("identifier") != null) {
|
199 |
52231
|
claudio.at
|
addField(instance, Instance.getDescriptor().findFieldByName("url"),
|
200 |
|
|
Lists.newArrayList(Iterables.filter(values.get("identifier").listValues(), urlFilter)));
|
201 |
42532
|
alessia.ba
|
}
|
202 |
26600
|
sandro.lab
|
|
203 |
42532
|
alessia.ba
|
result.addInstance(instance);
|
204 |
26600
|
sandro.lab
|
|
205 |
42532
|
alessia.ba
|
final List<Element> extrefs = values.get("reference");
|
206 |
|
|
if (!extrefs.isEmpty()) {
|
207 |
|
|
final Descriptor extDesc = ExternalReference.getDescriptor();
|
208 |
|
|
for (final Element element : extrefs) {
|
209 |
|
|
final ExternalReference.Builder extref = ExternalReference.newBuilder();
|
210 |
|
|
addField(extref, extDesc.findFieldByName("url"), element.getText());
|
211 |
|
|
addField(extref, extDesc.findFieldByName("sitename"), element.getAttributes().get("source"));
|
212 |
|
|
addField(extref, extDesc.findFieldByName("refidentifier"), element.getAttributes().get("identifier"));
|
213 |
|
|
addField(extref, extDesc.findFieldByName("label"), element.getAttributes().get("title"));
|
214 |
|
|
addField(extref, extDesc.findFieldByName("query"), element.getAttributes().get("query"));
|
215 |
|
|
addField(extref, extDesc.findFieldByName("qualifier"),
|
216 |
|
|
setQualifier(getDefaultQualifier("dnet:externalReference_typologies"), Lists.newArrayList(element.getAttributes().get("type")))
|
217 |
|
|
.build());
|
218 |
26600
|
sandro.lab
|
|
219 |
42532
|
alessia.ba
|
result.addExternalReference(extref);
|
220 |
|
|
}
|
221 |
|
|
}
|
222 |
38972
|
claudio.at
|
|
223 |
42532
|
alessia.ba
|
return result.setMetadata(metadata);
|
224 |
26600
|
sandro.lab
|
}
|
225 |
|
|
|
226 |
52169
|
claudio.at
|
private static void handleException(Throwable e, final String resultId, final ValueMap values) {
|
227 |
42532
|
alessia.ba
|
System.err.println("resultId: " + resultId);
|
228 |
52169
|
claudio.at
|
if (values != null) {
|
229 |
|
|
System.err.println("values: " + values);
|
230 |
|
|
}
|
231 |
42532
|
alessia.ba
|
e.printStackTrace();
|
232 |
|
|
throw new RuntimeException(e);
|
233 |
|
|
}
|
234 |
26600
|
sandro.lab
|
}
|