|
1
|
package eu.dnetlib.data.transform.xml;
|
|
2
|
|
|
3
|
import java.util.HashMap;
|
|
4
|
import java.util.List;
|
|
5
|
import java.util.Map;
|
|
6
|
import java.util.Set;
|
|
7
|
|
|
8
|
import com.google.common.collect.Lists;
|
|
9
|
import com.google.common.collect.Maps;
|
|
10
|
import com.google.common.collect.Sets;
|
|
11
|
import eu.dnetlib.data.mapreduce.util.OafRowKeyDecoder;
|
|
12
|
import eu.dnetlib.data.proto.FieldTypeProtos.Author;
|
|
13
|
import eu.dnetlib.data.proto.FieldTypeProtos.KeyValue;
|
|
14
|
import eu.dnetlib.data.proto.FieldTypeProtos.StructuredProperty;
|
|
15
|
import eu.dnetlib.data.proto.OafProtos.Oaf;
|
|
16
|
import eu.dnetlib.data.proto.OafProtos.OafEntity;
|
|
17
|
import eu.dnetlib.data.proto.ResultProtos.Result;
|
|
18
|
import eu.dnetlib.data.proto.ResultProtos.Result.Context;
|
|
19
|
import eu.dnetlib.data.proto.ResultProtos.Result.Instance;
|
|
20
|
import eu.dnetlib.data.proto.TypeProtos.Type;
|
|
21
|
import org.apache.commons.lang3.StringUtils;
|
|
22
|
import org.w3c.dom.Element;
|
|
23
|
import org.w3c.dom.NamedNodeMap;
|
|
24
|
import org.w3c.dom.Node;
|
|
25
|
import org.w3c.dom.NodeList;
|
|
26
|
|
|
27
|
public class OdfToHbaseXsltFunctions extends CommonDNetXsltFunctions {
|
|
28
|
|
|
29
|
private static Map<String, String> mappingAccess = Maps.newHashMap();
|
|
30
|
|
|
31
|
static {
|
|
32
|
|
|
33
|
mappingAccess.put("info:eu-repo/semantics/openAccess", "OPEN");
|
|
34
|
mappingAccess.put("info:eu-repo/semantics/closedAccess", "CLOSED");
|
|
35
|
mappingAccess.put("info:eu-repo/semantics/restrictedAccess", "RESTRICTED");
|
|
36
|
mappingAccess.put("info:eu-repo/semantics/embargoedAccess", "EMBARGO");
|
|
37
|
|
|
38
|
// Transformator now maps the access rights into proper values, not sure if it does for all datasets.
|
|
39
|
mappingAccess.put("OPEN", "OPEN");
|
|
40
|
mappingAccess.put("CLOSED", "CLOSED");
|
|
41
|
mappingAccess.put("RESTRICTED", "RESTRICTED");
|
|
42
|
mappingAccess.put("EMBARGO", "EMBARGO");
|
|
43
|
mappingAccess.put("OPEN SOURCE", "OPEN SOURCE");
|
|
44
|
|
|
45
|
}
|
|
46
|
|
|
47
|
public static String odfResult(
|
|
48
|
final String resultId,
|
|
49
|
final boolean invisible,
|
|
50
|
final NodeList about,
|
|
51
|
final NodeList metadata,
|
|
52
|
final NodeList titles,
|
|
53
|
final NodeList creators,
|
|
54
|
final NodeList subjects,
|
|
55
|
final NodeList publisher,
|
|
56
|
final NodeList descriptions,
|
|
57
|
final NodeList dates,
|
|
58
|
final NodeList dateaccepted,
|
|
59
|
final NodeList resourceTypes,
|
|
60
|
final NodeList formats,
|
|
61
|
final NodeList sizes,
|
|
62
|
final NodeList languages,
|
|
63
|
final NodeList cobjcategory,
|
|
64
|
final NodeList contributors,
|
|
65
|
final NodeList rights,
|
|
66
|
final NodeList license,
|
|
67
|
final NodeList version,
|
|
68
|
final NodeList pidList,
|
|
69
|
final String provenance,
|
|
70
|
final String trust,
|
|
71
|
final NodeList hostedby,
|
|
72
|
final NodeList collectedfrom,
|
|
73
|
final NodeList originalIds,
|
|
74
|
final String instanceUri,
|
|
75
|
final String landingPage,
|
|
76
|
final NodeList distributionlocation,
|
|
77
|
final NodeList documentationUrl,
|
|
78
|
final String dateOfCollection,
|
|
79
|
final String dateOfTransformation) {
|
|
80
|
|
|
81
|
try {
|
|
82
|
final String entityId = OafRowKeyDecoder.decode(resultId).getKey();
|
|
83
|
|
|
84
|
final Result.Builder result = Result.newBuilder();
|
|
85
|
Result.Metadata.Builder metadataProto = Result.Metadata.newBuilder();
|
|
86
|
|
|
87
|
// subject
|
|
88
|
for (int i = 0; i < subjects.getLength(); i++) {
|
|
89
|
Node currentNode = subjects.item(i);
|
|
90
|
NodeList childNodes = currentNode.getChildNodes();
|
|
91
|
if (childNodes.getLength() > 0) {
|
|
92
|
String subjectValue = childNodes.item(0).getNodeValue();
|
|
93
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("subject"),
|
|
94
|
getStructuredProperty(subjectValue, "keyword", "keyword", "dnet:subject_classification_typologies", "dnet:subject_classification_typologies"));
|
|
95
|
}
|
|
96
|
}
|
|
97
|
|
|
98
|
// title
|
|
99
|
for (int i = 0; i < titles.getLength(); i++) {
|
|
100
|
Node currentNode = titles.item(i);
|
|
101
|
NodeList childNodes = currentNode.getChildNodes();
|
|
102
|
if (childNodes.getLength() > 0) {
|
|
103
|
String titleValue = childNodes.item(0).getNodeValue();
|
|
104
|
String classname = "main title";
|
|
105
|
String classid = "main title";
|
|
106
|
if (currentNode.hasAttributes()) {
|
|
107
|
NamedNodeMap attributes = currentNode.getAttributes();
|
|
108
|
Node titleType = attributes.getNamedItem("titleType");
|
|
109
|
|
|
110
|
if (titleType != null && titleType.getNodeValue().equals("AlternativeTitle")) {
|
|
111
|
classname = "alternative title";
|
|
112
|
classid = "alternative title";
|
|
113
|
}
|
|
114
|
if (titleType != null && titleType.getNodeValue().equals("Subtitle")) {
|
|
115
|
classname = "subtitle";
|
|
116
|
classid = "subtitle";
|
|
117
|
}
|
|
118
|
if (titleType != null && titleType.getNodeValue().equals("TranslatedTitle")) {
|
|
119
|
classname = "translated title";
|
|
120
|
classid = "translated title";
|
|
121
|
}
|
|
122
|
}
|
|
123
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("title"),
|
|
124
|
getStructuredProperty(titleValue, classname, classid, "dnet:dataCite_title", "dnet:dataCite_title"));
|
|
125
|
}
|
|
126
|
}
|
|
127
|
|
|
128
|
// creators
|
|
129
|
for (int i = 0; i < creators.getLength(); i++) {
|
|
130
|
final Element creator = (Element) creators.item(i);
|
|
131
|
if (creator != null && creator.hasChildNodes()) {
|
|
132
|
|
|
133
|
final NodeList creatorNames = creator.getElementsByTagName("creatorName");
|
|
134
|
if (creatorNames.getLength() > 0) {
|
|
135
|
final Element creatorName = (Element) creatorNames.item(0);
|
|
136
|
|
|
137
|
final Author.Builder author = Author.newBuilder();
|
|
138
|
author.setRank(i+1);
|
|
139
|
final String fullname = StringUtils.trim(creatorName.getTextContent());
|
|
140
|
|
|
141
|
author.setFullname(fullname);
|
|
142
|
|
|
143
|
final eu.dnetlib.pace.model.Person p = new eu.dnetlib.pace.model.Person(fullname, false);
|
|
144
|
if (p.isAccurate()) {
|
|
145
|
author.setName(p.getNormalisedFirstName());
|
|
146
|
author.setSurname(p.getNormalisedSurname());
|
|
147
|
}
|
|
148
|
final NodeList nameIdentifiers = creator.getElementsByTagName("nameIdentifier");
|
|
149
|
if (nameIdentifiers.getLength() > 0) {
|
|
150
|
final Element nameIdentifier = (Element) nameIdentifiers.item(0);
|
|
151
|
final String nameIdentifierScheme = nameIdentifier.getAttribute("nameIdentifierScheme");
|
|
152
|
final String id = StringUtils.trim(nameIdentifier.getTextContent());
|
|
153
|
if (StringUtils.isNotBlank(id) && StringUtils.isNotBlank(nameIdentifierScheme)) {
|
|
154
|
author.addPid(getKV(nameIdentifierScheme, id));
|
|
155
|
}
|
|
156
|
}
|
|
157
|
|
|
158
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("author"), author);
|
|
159
|
}
|
|
160
|
}
|
|
161
|
}
|
|
162
|
|
|
163
|
// description
|
|
164
|
for (int i = 0; i < descriptions.getLength(); i++) {
|
|
165
|
Element currentNode = (Element) descriptions.item(i);
|
|
166
|
if (currentNode != null && currentNode.hasChildNodes()) {
|
|
167
|
String descriptionValue = currentNode.getChildNodes().item(0).getNodeValue();
|
|
168
|
|
|
169
|
final String descriptionType = currentNode.getAttribute("descriptionType");
|
|
170
|
if (StringUtils.isNotBlank(descriptionType)) {
|
|
171
|
switch (descriptionType) {
|
|
172
|
case "TechnicalInfo":
|
|
173
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("tool"), descriptionValue);
|
|
174
|
break;
|
|
175
|
case "Abstract":
|
|
176
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("description"), descriptionValue);
|
|
177
|
break;
|
|
178
|
case "DistributionForm":
|
|
179
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("format"), descriptionValue);
|
|
180
|
break;
|
|
181
|
}
|
|
182
|
} else {
|
|
183
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("description"), descriptionValue);
|
|
184
|
}
|
|
185
|
}
|
|
186
|
}
|
|
187
|
|
|
188
|
// contributors
|
|
189
|
for (int i = 0; i < contributors.getLength(); i++) {
|
|
190
|
final Element contributor = (Element) contributors.item(i);
|
|
191
|
if (contributor != null && contributor.hasChildNodes()) {
|
|
192
|
|
|
193
|
NodeList contributorNames = contributor.getElementsByTagName("contributorName");
|
|
194
|
if (contributorNames != null) {
|
|
195
|
Element contributorName = (Element) contributorNames.item(0);
|
|
196
|
if (contributorName != null) {
|
|
197
|
final String contributorValue = contributorName.getTextContent();
|
|
198
|
final String contributorType = contributor.getAttribute("contributorType");
|
|
199
|
|
|
200
|
if (StringUtils.isNotBlank(contributorType)) {
|
|
201
|
switch (contributorType) {
|
|
202
|
case "ContactPerson":
|
|
203
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("contactperson"), contributorValue);
|
|
204
|
break;
|
|
205
|
case "ContactGroup":
|
|
206
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("contactgroup"), contributorValue);
|
|
207
|
break;
|
|
208
|
}
|
|
209
|
} else {
|
|
210
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("contributor"), contributorValue);
|
|
211
|
}
|
|
212
|
}
|
|
213
|
}
|
|
214
|
}
|
|
215
|
}
|
|
216
|
|
|
217
|
// publisher
|
|
218
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("publisher"), getFirstItem(publisher));
|
|
219
|
|
|
220
|
// dates
|
|
221
|
for (int i = 0; i < dates.getLength(); i++) {
|
|
222
|
Node currentNode = dates.item(i);
|
|
223
|
if (currentNode != null && currentNode.hasAttributes() && currentNode.hasChildNodes()) {
|
|
224
|
String dateAttribute = currentNode.getAttributes().getNamedItem("dateType").getNodeValue();
|
|
225
|
String dateValue = currentNode.getChildNodes().item(0).getNodeValue();
|
|
226
|
String protoAttribute = "relevantdate";
|
|
227
|
if ("Accepted".equals(dateAttribute)) {
|
|
228
|
protoAttribute = "dateofacceptance";
|
|
229
|
} else if ("Issued".equals(dateAttribute)) {
|
|
230
|
protoAttribute = "storagedate";
|
|
231
|
} else if ("Updated".equals(dateAttribute)) {
|
|
232
|
protoAttribute = "lastmetadataupdate";
|
|
233
|
} else if ("Available".equals(dateAttribute)) {
|
|
234
|
protoAttribute = "embargoenddate";
|
|
235
|
}
|
|
236
|
if (protoAttribute.equals("relevantdate") == false) {
|
|
237
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName(protoAttribute), dateValue);
|
|
238
|
} else {
|
|
239
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName(protoAttribute),
|
|
240
|
getStructuredProperty(dateValue, "UNKNOWN", "UNKNOWN", "dnet:dataCite_date", "dnet:dataCite_date"));
|
|
241
|
}
|
|
242
|
}
|
|
243
|
}
|
|
244
|
|
|
245
|
// dateofacceptance
|
|
246
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("dateofacceptance"), getFirstItem(dateaccepted));
|
|
247
|
|
|
248
|
// size
|
|
249
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("size"), getFirstItem(sizes));
|
|
250
|
|
|
251
|
// version
|
|
252
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("version"), getFirstItem(version));
|
|
253
|
|
|
254
|
// language
|
|
255
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("language"),
|
|
256
|
setQualifier(getDefaultQualifier("dnet:languages"), Lists.newArrayList(getFirstItem(languages))));
|
|
257
|
|
|
258
|
// resource type
|
|
259
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("resourcetype"),
|
|
260
|
setQualifier(getDefaultQualifier("dnet:dataCite_resource"), Lists.newArrayList(getFirstItem(resourceTypes))));
|
|
261
|
|
|
262
|
// resultType
|
|
263
|
final String cobjcategoryCode = getFirstItem(cobjcategory);
|
|
264
|
final String resulttype = getResultType(cobjcategory);
|
|
265
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("resulttype"), getSimpleQualifier(resulttype, "dnet:result_typologies"));
|
|
266
|
|
|
267
|
switch (resulttype) {
|
|
268
|
case "software" :
|
|
269
|
// format
|
|
270
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("programmingLanguage"),
|
|
271
|
getSimpleQualifier(getFirstItem(formats), "dnet:programming_languages"));
|
|
272
|
break;
|
|
273
|
case "dataset":
|
|
274
|
for (int i = 0; i < formats.getLength(); i++) {
|
|
275
|
Node currentNode = formats.item(i);
|
|
276
|
NodeList childNodes = currentNode.getChildNodes();
|
|
277
|
if (childNodes.getLength() > 0) {
|
|
278
|
String formatValue = childNodes.item(0).getNodeValue();
|
|
279
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("format"), formatValue);
|
|
280
|
}
|
|
281
|
}
|
|
282
|
break;
|
|
283
|
case "other":
|
|
284
|
|
|
285
|
break;
|
|
286
|
}
|
|
287
|
|
|
288
|
// documentationUrl
|
|
289
|
for (int i = 0; i < documentationUrl.getLength(); i++) {
|
|
290
|
final Element docUrl = (Element) documentationUrl.item(i);
|
|
291
|
if (docUrl != null && docUrl.hasChildNodes()) {
|
|
292
|
final String value = docUrl.getTextContent();
|
|
293
|
if (StringUtils.isNotBlank(value)) {
|
|
294
|
addField(metadataProto, Result.Metadata.getDescriptor().findFieldByName("documentationUrl"), value);
|
|
295
|
}
|
|
296
|
}
|
|
297
|
}
|
|
298
|
|
|
299
|
// contexts
|
|
300
|
ValueMap values = ValueMap.parseNodeList(metadata);
|
|
301
|
if (values.get("concept") != null) {
|
|
302
|
for (final eu.dnetlib.data.transform.xml.Element e : values.get("concept")) {
|
|
303
|
final String id = e.getAttributes().get("id");
|
|
304
|
if (StringUtils.isBlank(id)) throw new IllegalArgumentException("Context id cannot be blank");
|
|
305
|
metadataProto.addContext(Context.newBuilder().setId(id));
|
|
306
|
}
|
|
307
|
}
|
|
308
|
|
|
309
|
final List<KeyValue> hostedBys = getKeyValues(ValueMap.parseNodeList(hostedby), "hostedby", Type.datasource);
|
|
310
|
final List<KeyValue> collectedFroms = getKeyValues(ValueMap.parseNodeList(collectedfrom), "collectedfrom", Type.datasource);
|
|
311
|
|
|
312
|
final Instance.Builder instance = Instance.newBuilder();
|
|
313
|
|
|
314
|
String tmpRigths = "UNKNOWN";
|
|
315
|
final String firstRight = getFirstItem(rights);
|
|
316
|
if (mappingAccess.containsKey(firstRight)) {
|
|
317
|
tmpRigths = mappingAccess.get(firstRight);
|
|
318
|
}
|
|
319
|
|
|
320
|
addField(instance, Instance.getDescriptor().findFieldByName("license"), getFirstItem(license));
|
|
321
|
addField(instance, Instance.getDescriptor().findFieldByName("hostedby"), hostedBys);
|
|
322
|
|
|
323
|
addField(instance, Instance.getDescriptor().findFieldByName("accessright"),
|
|
324
|
setQualifier(getDefaultQualifier("dnet:access_modes"), Lists.newArrayList(tmpRigths)));
|
|
325
|
|
|
326
|
addField(instance, Instance.getDescriptor().findFieldByName("instancetype"),
|
|
327
|
setQualifier(getDefaultQualifier("dnet:dataCite_resource"), Lists.newArrayList(cobjcategoryCode)));
|
|
328
|
|
|
329
|
addField(instance, Instance.getDescriptor().findFieldByName("url"), instanceUri);
|
|
330
|
if (StringUtils.isNotBlank(landingPage)) {
|
|
331
|
addField(instance, Instance.getDescriptor().findFieldByName("url"), landingPage);
|
|
332
|
}
|
|
333
|
addField(instance, Instance.getDescriptor().findFieldByName("distributionlocation"), getFirstItem(distributionlocation));
|
|
334
|
|
|
335
|
addField(instance, Instance.getDescriptor().findFieldByName("collectedfrom"), collectedFroms);
|
|
336
|
addField(instance, Instance.getDescriptor().findFieldByName("dateofacceptance"), getFirstItem(dateaccepted));
|
|
337
|
|
|
338
|
result.addInstance(instance);
|
|
339
|
|
|
340
|
List<StructuredProperty> pids = parsePids(pidList);
|
|
341
|
|
|
342
|
// original ids
|
|
343
|
final Set<String> originalIdList = Sets.newHashSet();
|
|
344
|
for (int i = 0; i < originalIds.getLength(); i++) {
|
|
345
|
Node currentNode = originalIds.item(i);
|
|
346
|
if (currentNode != null && currentNode.hasChildNodes()) {
|
|
347
|
originalIdList.add(currentNode.getChildNodes().item(0).getNodeValue());
|
|
348
|
}
|
|
349
|
}
|
|
350
|
|
|
351
|
OafEntity.Builder entity =
|
|
352
|
getEntity(Type.result, entityId, collectedFroms, originalIdList, dateOfCollection, dateOfTransformation, pids).setResult(
|
|
353
|
result.setMetadata(metadataProto));
|
|
354
|
|
|
355
|
entity.setOaiprovenance(getOAIProvenance(about));
|
|
356
|
|
|
357
|
Oaf oaf = getOaf(entity, getDataInfo(invisible, about, provenance, trust, false, false));
|
|
358
|
return base64(oaf.toByteArray());
|
|
359
|
} catch (Exception e) {
|
|
360
|
e.printStackTrace(System.err);
|
|
361
|
throw new RuntimeException(e);
|
|
362
|
}
|
|
363
|
|
|
364
|
}
|
|
365
|
|
|
366
|
private static String getResultType(final NodeList cobjcategoryNode) {
|
|
367
|
|
|
368
|
final ValueMap values = ValueMap.parseNodeList(cobjcategoryNode);
|
|
369
|
|
|
370
|
final eu.dnetlib.data.transform.xml.Element cobjcategory = values.get("cobjcategory").stream()
|
|
371
|
.map(e -> StringUtils.isNotBlank(e.getText()) ? e : new eu.dnetlib.data.transform.xml.Element("0000", e.getAttributes()))
|
|
372
|
.findFirst()
|
|
373
|
.orElse(new eu.dnetlib.data.transform.xml.Element("0000", new HashMap<>()));
|
|
374
|
|
|
375
|
final String resulttype = cobjcategory.getAttributeValue("type");
|
|
376
|
if (StringUtils.isNotBlank(resulttype)) {
|
|
377
|
return resulttype;
|
|
378
|
}
|
|
379
|
|
|
380
|
return getDefaultResulttype(cobjcategory);
|
|
381
|
}
|
|
382
|
|
|
383
|
public static String getFirstItem(final NodeList list) {
|
|
384
|
String out = "";
|
|
385
|
if (list != null) {
|
|
386
|
|
|
387
|
if (list.getLength() > 0 && list.item(0).getChildNodes() != null && list.item(0).getChildNodes().getLength() > 0) {
|
|
388
|
out = list.item(0).getChildNodes().item(0).getNodeValue();
|
|
389
|
}
|
|
390
|
}
|
|
391
|
return out;
|
|
392
|
}
|
|
393
|
|
|
394
|
}
|