1 |
27199
|
claudio.at
|
package eu.dnetlib.data.mapreduce.hbase.oai;
|
2 |
|
|
|
3 |
|
|
import java.io.IOException;
|
4 |
28303
|
alessia.ba
|
import java.net.UnknownHostException;
|
5 |
52078
|
alessia.ba
|
import java.time.LocalDateTime;
|
6 |
|
|
import java.time.ZoneId;
|
7 |
|
|
import java.time.format.DateTimeFormatter;
|
8 |
28303
|
alessia.ba
|
import java.util.Collection;
|
9 |
|
|
import java.util.Date;
|
10 |
29014
|
alessia.ba
|
import java.util.Map;
|
11 |
38665
|
alessia.ba
|
import java.util.zip.ZipEntry;
|
12 |
|
|
import java.util.zip.ZipOutputStream;
|
13 |
27199
|
claudio.at
|
|
14 |
52109
|
alessia.ba
|
import com.google.common.base.Splitter;
|
15 |
28303
|
alessia.ba
|
import com.google.common.collect.Iterables;
|
16 |
|
|
import com.google.common.collect.Lists;
|
17 |
29014
|
alessia.ba
|
import com.google.common.collect.Maps;
|
18 |
28303
|
alessia.ba
|
import com.google.common.collect.Multimap;
|
19 |
48892
|
claudio.at
|
import com.mongodb.BasicDBObject;
|
20 |
|
|
import com.mongodb.DBObject;
|
21 |
|
|
import com.mongodb.MongoClient;
|
22 |
|
|
import com.mongodb.WriteConcern;
|
23 |
45299
|
alessia.ba
|
import com.mongodb.client.MongoCollection;
|
24 |
|
|
import com.mongodb.client.MongoDatabase;
|
25 |
28303
|
alessia.ba
|
import eu.dnetlib.data.mapreduce.JobParams;
|
26 |
29000
|
alessia.ba
|
import eu.dnetlib.data.mapreduce.hbase.oai.config.OAIConfiguration;
|
27 |
|
|
import eu.dnetlib.data.mapreduce.hbase.oai.config.OAIConfigurationReader;
|
28 |
|
|
import eu.dnetlib.data.mapreduce.hbase.oai.config.OAIConfigurationStringReader;
|
29 |
|
|
import eu.dnetlib.data.mapreduce.hbase.oai.utils.MongoSetCollection;
|
30 |
|
|
import eu.dnetlib.data.mapreduce.hbase.oai.utils.PublisherField;
|
31 |
29014
|
alessia.ba
|
import eu.dnetlib.data.mapreduce.hbase.oai.utils.RecordFieldsExtractor;
|
32 |
29030
|
alessia.ba
|
import eu.dnetlib.miscutils.functional.xml.IndentXmlString;
|
33 |
45299
|
alessia.ba
|
import org.apache.commons.io.output.ByteArrayOutputStream;
|
34 |
|
|
import org.apache.commons.lang.StringUtils;
|
35 |
|
|
import org.apache.hadoop.io.NullWritable;
|
36 |
|
|
import org.apache.hadoop.io.Text;
|
37 |
|
|
import org.apache.hadoop.mapreduce.Mapper;
|
38 |
|
|
import org.bson.types.Binary;
|
39 |
28303
|
alessia.ba
|
|
40 |
27199
|
claudio.at
|
public class OaiFeedMapper extends Mapper<Text, Text, NullWritable, NullWritable> {
|
41 |
|
|
|
42 |
28303
|
alessia.ba
|
enum RecordStatus {
|
43 |
|
|
NEW, UPDATED, UNCHANGED;
|
44 |
|
|
}
|
45 |
|
|
|
46 |
45299
|
alessia.ba
|
private MongoCollection<DBObject> collection;
|
47 |
|
|
private MongoCollection<DBObject> discardedCollection;
|
48 |
28303
|
alessia.ba
|
private OAIConfigurationStringReader oaiConfigurationReader;
|
49 |
|
|
private OAIConfiguration oaiConfiguration;
|
50 |
27199
|
claudio.at
|
|
51 |
28303
|
alessia.ba
|
private Date feedDate;
|
52 |
|
|
|
53 |
|
|
private MongoSetCollection mongoSetCollection;
|
54 |
|
|
|
55 |
29014
|
alessia.ba
|
private RecordFieldsExtractor extractor;
|
56 |
28693
|
claudio.at
|
|
57 |
28904
|
alessia.ba
|
// these are set in the setup
|
58 |
|
|
private String format;
|
59 |
|
|
private String interpretation;
|
60 |
|
|
private String layout;
|
61 |
29014
|
alessia.ba
|
private Map<String, PublisherField> fieldsToIndex = Maps.newHashMap();
|
62 |
28904
|
alessia.ba
|
|
63 |
34374
|
alessia.ba
|
private String duplicateXPath;
|
64 |
|
|
private boolean skipDuplicates;
|
65 |
|
|
|
66 |
37794
|
alessia.ba
|
private MongoClient mongo;
|
67 |
|
|
|
68 |
37821
|
alessia.ba
|
private Collection<String> enrichmentXPaths;
|
69 |
|
|
|
70 |
52109
|
alessia.ba
|
private String[] parseDatePatterns;
|
71 |
|
|
|
72 |
27199
|
claudio.at
|
@Override
|
73 |
28303
|
alessia.ba
|
protected void setup(final Context context) throws UnknownHostException {
|
74 |
27199
|
claudio.at
|
|
75 |
|
|
String host = context.getConfiguration().get("services.publisher.oai.host");
|
76 |
|
|
String port = context.getConfiguration().get("services.publisher.oai.port");
|
77 |
|
|
String db = context.getConfiguration().get("services.publisher.oai.db");
|
78 |
|
|
String collectionName = context.getConfiguration().get("services.publisher.oai.collection");
|
79 |
52109
|
alessia.ba
|
String patterns = context.getConfiguration().get("services.publisher.oai.datepatterns");
|
80 |
52111
|
alessia.ba
|
this.parseDatePatterns = Lists.newArrayList(Splitter.on(',')
|
81 |
52109
|
alessia.ba
|
.trimResults()
|
82 |
52111
|
alessia.ba
|
.omitEmptyStrings().split(patterns)).toArray(new String[0]);
|
83 |
27199
|
claudio.at
|
|
84 |
|
|
System.out.println("Mongodb client params");
|
85 |
|
|
System.out.println("host: " + host);
|
86 |
|
|
System.out.println("port: " + port);
|
87 |
|
|
System.out.println("db: " + db);
|
88 |
|
|
System.out.println("collection: " + collectionName);
|
89 |
52109
|
alessia.ba
|
System.out.println("split date patterns: " + patterns);
|
90 |
27199
|
claudio.at
|
|
91 |
29041
|
claudio.at
|
String[] formatLayoutInterp = collectionName.split("-");
|
92 |
|
|
format = formatLayoutInterp[0];
|
93 |
|
|
layout = formatLayoutInterp[1];
|
94 |
|
|
interpretation = formatLayoutInterp[2];
|
95 |
28904
|
alessia.ba
|
|
96 |
28303
|
alessia.ba
|
String oaiConfigurationProfile = context.getConfiguration().get("oaiConfiguration");
|
97 |
29030
|
alessia.ba
|
System.out.println("oaiConfiguration:\n" + IndentXmlString.apply(oaiConfigurationProfile));
|
98 |
29000
|
alessia.ba
|
oaiConfigurationReader = new OAIConfigurationStringReader(oaiConfigurationProfile);
|
99 |
28303
|
alessia.ba
|
oaiConfiguration = oaiConfigurationReader.getOaiConfiguration();
|
100 |
|
|
|
101 |
29041
|
claudio.at
|
System.out.println("parsed configuration:" + oaiConfiguration.toString());
|
102 |
|
|
|
103 |
37794
|
alessia.ba
|
mongo = new MongoClient(host, Integer.parseInt(port));
|
104 |
45299
|
alessia.ba
|
MongoDatabase mongoDB = mongo.getDatabase(db);
|
105 |
52111
|
alessia.ba
|
|
106 |
45299
|
alessia.ba
|
collection = mongoDB.getCollection(collectionName, DBObject.class).withWriteConcern(WriteConcern.UNACKNOWLEDGED);
|
107 |
|
|
discardedCollection = mongoDB.getCollection("discarded-" + collectionName, DBObject.class).withWriteConcern(WriteConcern.UNACKNOWLEDGED);
|
108 |
|
|
mongoSetCollection = new MongoSetCollection(mongo);
|
109 |
34374
|
alessia.ba
|
|
110 |
|
|
duplicateXPath = context.getConfiguration().get("services.publisher.oai.duplicateXPath");
|
111 |
|
|
skipDuplicates = Boolean.parseBoolean(context.getConfiguration().get("services.publisher.oai.skipDuplicates"));
|
112 |
|
|
|
113 |
37821
|
alessia.ba
|
enrichmentXPaths = oaiConfiguration.getEnrichmentXPathsFor(format, layout, interpretation);
|
114 |
29014
|
alessia.ba
|
Collection<PublisherField> indexFields = oaiConfiguration.getFieldsFor(format, layout, interpretation);
|
115 |
|
|
extractor = new RecordFieldsExtractor(Lists.newArrayList(indexFields));
|
116 |
34374
|
alessia.ba
|
extractor.setDuplicateXPath(duplicateXPath);
|
117 |
|
|
extractor.setSkipDuplicates(skipDuplicates);
|
118 |
|
|
|
119 |
29014
|
alessia.ba
|
for (PublisherField field : indexFields) {
|
120 |
|
|
fieldsToIndex.put(field.getFieldName(), field);
|
121 |
|
|
}
|
122 |
27199
|
claudio.at
|
|
123 |
28303
|
alessia.ba
|
String feedDateString = context.getConfiguration().get(JobParams.OAI_FEED_DATE);
|
124 |
52076
|
alessia.ba
|
feedDate = parseDate(feedDateString);
|
125 |
27199
|
claudio.at
|
}
|
126 |
|
|
|
127 |
|
|
@Override
|
128 |
28303
|
alessia.ba
|
protected void map(final Text key, final Text value, final Context context) throws IOException, InterruptedException {
|
129 |
|
|
String recordKey = key.toString();
|
130 |
49029
|
claudio.at
|
String recordBody = value.toString();
|
131 |
|
|
if (StringUtils.isBlank(recordBody)) {
|
132 |
|
|
discard(context, recordKey, recordBody, "blank body");
|
133 |
|
|
} else {
|
134 |
|
|
Multimap<String, String> recordFields = extractor.extractFields(recordBody, enrichmentXPaths);
|
135 |
|
|
String id;
|
136 |
|
|
String oaiID;
|
137 |
|
|
if (checkRecordFields(recordFields, context, recordKey, recordBody)) {
|
138 |
|
|
id = recordFields.get(OAIConfigurationReader.ID_FIELD).iterator().next();
|
139 |
|
|
oaiID = getOAIIdentifier(id);
|
140 |
|
|
handleRecord(context, oaiID, recordBody, recordFields);
|
141 |
28303
|
alessia.ba
|
}
|
142 |
|
|
}
|
143 |
|
|
}
|
144 |
27199
|
claudio.at
|
|
145 |
34374
|
alessia.ba
|
public boolean checkRecordFields(final Multimap<String, String> recordFields, final Context context, final String recordKey, final String recordBody) {
|
146 |
|
|
if ((recordFields == null)) {
|
147 |
|
|
context.getCounter("oai", "invalid").increment(1);
|
148 |
|
|
return false;
|
149 |
|
|
}
|
150 |
|
|
if (recordFields.containsEntry("duplicate", "true")) {
|
151 |
|
|
if (skipDuplicates) {
|
152 |
|
|
context.getCounter("oai", "discardedDuplicate").increment(1);
|
153 |
|
|
return false;
|
154 |
|
|
} else return true;
|
155 |
|
|
}
|
156 |
|
|
if (!recordFields.containsKey(OAIConfigurationReader.ID_FIELD)) {
|
157 |
|
|
discard(context, recordKey, recordBody, "missing " + OAIConfigurationReader.ID_FIELD);
|
158 |
|
|
return false;
|
159 |
|
|
}
|
160 |
|
|
return true;
|
161 |
|
|
}
|
162 |
|
|
|
163 |
38671
|
alessia.ba
|
private void handleRecord(final Context context, final String oaiID, final String record, final Multimap<String, String> recordProperties) {
|
164 |
45299
|
alessia.ba
|
DBObject obj = this.createBasicObject(oaiID, record, recordProperties,context);
|
165 |
38671
|
alessia.ba
|
if (obj != null) { // it can be null if the compression did not succeeded: counter is updated in the compress method in that case
|
166 |
49891
|
alessia.ba
|
//let's use the date of collection and transformation for those records that have them.
|
167 |
|
|
//for the rest, we'll keep the feedDate...
|
168 |
|
|
Collection<String> collectionDates = recordProperties.get("dateOfCollection");
|
169 |
|
|
Collection<String> transDates = recordProperties.get("dateOfTransformation");
|
170 |
52076
|
alessia.ba
|
Date collDate = feedDate;
|
171 |
49891
|
alessia.ba
|
if(collectionDates != null && !collectionDates.isEmpty()){
|
172 |
52076
|
alessia.ba
|
String collDateString = collectionDates.iterator().next();
|
173 |
|
|
if(StringUtils.isNotBlank(collDateString))
|
174 |
|
|
collDate = parseDate(collDateString);
|
175 |
49891
|
alessia.ba
|
}
|
176 |
52114
|
alessia.ba
|
else{
|
177 |
|
|
context.getCounter("oai", "missing dateOfCollection").increment(1);
|
178 |
|
|
}
|
179 |
52076
|
alessia.ba
|
obj.put(OAIConfigurationReader.LAST_COLLECTION_DATE_FIELD, collDate);
|
180 |
|
|
Date timestamp = feedDate;
|
181 |
49891
|
alessia.ba
|
if(transDates != null && !transDates.isEmpty()){
|
182 |
52076
|
alessia.ba
|
String transDateString = transDates.iterator().next();
|
183 |
|
|
if(StringUtils.isNotBlank(transDateString))
|
184 |
|
|
timestamp = parseDate(transDateString);
|
185 |
49891
|
alessia.ba
|
}
|
186 |
52114
|
alessia.ba
|
else{
|
187 |
|
|
context.getCounter("oai", "missing dateOfTransformation").increment(1);
|
188 |
|
|
}
|
189 |
52076
|
alessia.ba
|
obj.put(OAIConfigurationReader.DATESTAMP_FIELD, timestamp);
|
190 |
49891
|
alessia.ba
|
//the updated_field must stay false, as the oai store is empty.
|
191 |
28303
|
alessia.ba
|
obj.put(OAIConfigurationReader.UPDATED_FIELD, false);
|
192 |
45299
|
alessia.ba
|
collection.insertOne(obj);
|
193 |
38665
|
alessia.ba
|
context.getCounter("oai", "total").increment(1);
|
194 |
|
|
}
|
195 |
28303
|
alessia.ba
|
}
|
196 |
27199
|
claudio.at
|
|
197 |
52078
|
alessia.ba
|
protected Date parseDate(final String date) {
|
198 |
|
|
// date should be in the form: 2017-12-18 or 2017-12-18T12:00:04+00:00 or 2014-10-12T15:54:02.43Z or 2014-10-12T15:54:02.432Z or 2014-10-12T15:54:02.4Z
|
199 |
52076
|
alessia.ba
|
try {
|
200 |
52078
|
alessia.ba
|
LocalDateTime d = LocalDateTime.parse(date, DateTimeFormatter.ISO_ZONED_DATE_TIME);
|
201 |
|
|
return Date.from(d.atZone(ZoneId.systemDefault()).toInstant());
|
202 |
|
|
}catch(Exception dateException){
|
203 |
|
|
try {
|
204 |
|
|
return org.apache.commons.lang.time.DateUtils.parseDate(
|
205 |
|
|
date,
|
206 |
52109
|
alessia.ba
|
parseDatePatterns);
|
207 |
52078
|
alessia.ba
|
}catch(Exception dateException2){
|
208 |
|
|
dateException2.printStackTrace(System.err);
|
209 |
|
|
throw new RuntimeException(dateException2);
|
210 |
|
|
}
|
211 |
52076
|
alessia.ba
|
}
|
212 |
|
|
}
|
213 |
45299
|
alessia.ba
|
|
214 |
28303
|
alessia.ba
|
private void discard(final Context context, final String recordKey, final String recordBody, final String reason) {
|
215 |
|
|
context.getCounter("oai", reason).increment(1);
|
216 |
45299
|
alessia.ba
|
discardedCollection.insertOne(new BasicDBObject("id", recordKey).append(OAIConfigurationReader.BODY_FIELD, recordBody));
|
217 |
28303
|
alessia.ba
|
}
|
218 |
|
|
|
219 |
|
|
private String getOAIIdentifier(final String id) {
|
220 |
29009
|
claudio.at
|
return oaiConfiguration.getIdScheme() + ":" + oaiConfiguration.getIdNamespace() + ":" + id;
|
221 |
28303
|
alessia.ba
|
}
|
222 |
|
|
|
223 |
29014
|
alessia.ba
|
protected DBObject createBasicObject(final String oaiID, final String record, final Multimap<String, String> recordProperties, final Context context) {
|
224 |
28303
|
alessia.ba
|
DBObject obj = new BasicDBObject();
|
225 |
|
|
for (final String key : recordProperties.keySet()) {
|
226 |
|
|
if (key.equals(OAIConfigurationReader.ID_FIELD)) {
|
227 |
|
|
obj.put(key, oaiID);
|
228 |
|
|
} else {
|
229 |
|
|
Collection<String> values = recordProperties.get(key);
|
230 |
|
|
if (key.equals(OAIConfigurationReader.SET_FIELD)) {
|
231 |
|
|
|
232 |
52112
|
alessia.ba
|
Iterable<String> setSpecs = Iterables.transform(values, s -> mongoSetCollection.normalizeSetSpec(s));
|
233 |
28303
|
alessia.ba
|
obj.put(key, setSpecs);
|
234 |
|
|
} else {
|
235 |
29014
|
alessia.ba
|
PublisherField keyField = fieldsToIndex.get(key);
|
236 |
52112
|
alessia.ba
|
//Skipping record properties that are not listed as fields to index (e.g. date of transformation and collection)
|
237 |
|
|
if (keyField != null) {
|
238 |
|
|
// let's check if the key is the name of a repeatable field or not
|
239 |
|
|
if ((keyField != null) && !keyField.isRepeatable()) {
|
240 |
|
|
if ((values != null) && !values.isEmpty()) {
|
241 |
|
|
obj.put(key, values.iterator().next());
|
242 |
|
|
}
|
243 |
|
|
} else {
|
244 |
|
|
obj.put(key, values);
|
245 |
28303
|
alessia.ba
|
}
|
246 |
|
|
}
|
247 |
|
|
}
|
248 |
|
|
}
|
249 |
|
|
}
|
250 |
38665
|
alessia.ba
|
|
251 |
|
|
Binary compressedRecordBody = createCompressRecord(context, oaiID, record);
|
252 |
|
|
if (compressedRecordBody != null) {
|
253 |
|
|
obj.put(OAIConfigurationReader.BODY_FIELD, compressedRecordBody);
|
254 |
|
|
obj.put(OAIConfigurationReader.DELETED_FIELD, false);
|
255 |
|
|
return obj;
|
256 |
|
|
} else return null;
|
257 |
28303
|
alessia.ba
|
}
|
258 |
|
|
|
259 |
38665
|
alessia.ba
|
public Binary createCompressRecord(final Context context, final String recordKey, final String recordBody) {
|
260 |
|
|
try {
|
261 |
|
|
ByteArrayOutputStream os = new ByteArrayOutputStream();
|
262 |
|
|
ZipOutputStream zos = new ZipOutputStream(os);
|
263 |
|
|
ZipEntry entry = new ZipEntry(OAIConfigurationReader.BODY_FIELD);
|
264 |
|
|
zos.putNextEntry(entry);
|
265 |
|
|
zos.write(recordBody.getBytes());
|
266 |
|
|
zos.closeEntry();
|
267 |
|
|
zos.flush();
|
268 |
|
|
zos.close();
|
269 |
|
|
return new Binary(os.toByteArray());
|
270 |
|
|
} catch (IOException e) {
|
271 |
|
|
discard(context, recordKey, recordBody, "cannot compress");
|
272 |
|
|
return null;
|
273 |
|
|
}
|
274 |
|
|
}
|
275 |
|
|
|
276 |
27199
|
claudio.at
|
@Override
|
277 |
28303
|
alessia.ba
|
protected void cleanup(final Context context) throws IOException, InterruptedException {
|
278 |
27199
|
claudio.at
|
|
279 |
|
|
super.cleanup(context);
|
280 |
|
|
}
|
281 |
|
|
|
282 |
45299
|
alessia.ba
|
public MongoCollection<DBObject> getCollection() {
|
283 |
28303
|
alessia.ba
|
return collection;
|
284 |
|
|
}
|
285 |
|
|
|
286 |
45299
|
alessia.ba
|
public void setCollection(final MongoCollection<DBObject> collection) {
|
287 |
28303
|
alessia.ba
|
this.collection = collection;
|
288 |
|
|
}
|
289 |
|
|
|
290 |
45299
|
alessia.ba
|
public MongoCollection<DBObject> getDiscardedCollection() {
|
291 |
28303
|
alessia.ba
|
return discardedCollection;
|
292 |
|
|
}
|
293 |
|
|
|
294 |
45299
|
alessia.ba
|
public void setDiscardedCollection(final MongoCollection<DBObject> discardedCollection) {
|
295 |
28303
|
alessia.ba
|
this.discardedCollection = discardedCollection;
|
296 |
|
|
}
|
297 |
|
|
|
298 |
|
|
public OAIConfigurationStringReader getOaiConfigurationReader() {
|
299 |
|
|
return oaiConfigurationReader;
|
300 |
|
|
}
|
301 |
|
|
|
302 |
|
|
public void setOaiConfigurationReader(final OAIConfigurationStringReader oaiConfigurationReader) {
|
303 |
|
|
this.oaiConfigurationReader = oaiConfigurationReader;
|
304 |
|
|
}
|
305 |
|
|
|
306 |
|
|
public OAIConfiguration getOaiConfiguration() {
|
307 |
|
|
return oaiConfiguration;
|
308 |
|
|
}
|
309 |
|
|
|
310 |
|
|
public void setOaiConfiguration(final OAIConfiguration oaiConfiguration) {
|
311 |
|
|
this.oaiConfiguration = oaiConfiguration;
|
312 |
|
|
}
|
313 |
|
|
|
314 |
|
|
public Date getFeedDate() {
|
315 |
|
|
return feedDate;
|
316 |
|
|
}
|
317 |
|
|
|
318 |
|
|
public void setFeedDate(final Date feedDate) {
|
319 |
|
|
this.feedDate = feedDate;
|
320 |
|
|
}
|
321 |
|
|
|
322 |
|
|
public MongoSetCollection getMongoSetCollection() {
|
323 |
|
|
return mongoSetCollection;
|
324 |
|
|
}
|
325 |
|
|
|
326 |
|
|
public void setMongoSetCollection(final MongoSetCollection mongoSetCollection) {
|
327 |
|
|
this.mongoSetCollection = mongoSetCollection;
|
328 |
|
|
}
|
329 |
|
|
|
330 |
34374
|
alessia.ba
|
public String getDuplicateXPath() {
|
331 |
|
|
return duplicateXPath;
|
332 |
|
|
}
|
333 |
|
|
|
334 |
|
|
public void setDuplicateXPath(final String duplicateXPath) {
|
335 |
|
|
this.duplicateXPath = duplicateXPath;
|
336 |
|
|
}
|
337 |
|
|
|
338 |
|
|
public boolean isSkipDuplicates() {
|
339 |
|
|
return skipDuplicates;
|
340 |
|
|
}
|
341 |
|
|
|
342 |
|
|
public void setSkipDuplicates(final boolean skipDuplicates) {
|
343 |
|
|
this.skipDuplicates = skipDuplicates;
|
344 |
|
|
}
|
345 |
|
|
|
346 |
52109
|
alessia.ba
|
public String[] getParseDatePatterns() {
|
347 |
|
|
return parseDatePatterns;
|
348 |
|
|
}
|
349 |
|
|
|
350 |
|
|
public void setParseDatePatterns(final String[] parseDatePatterns) {
|
351 |
|
|
this.parseDatePatterns = parseDatePatterns;
|
352 |
|
|
}
|
353 |
27199
|
claudio.at
|
}
|