Project

General

Profile

1
package eu.dnetlib.data.mapreduce.hbase.oai;
2

    
3
import java.io.IOException;
4
import java.net.UnknownHostException;
5
import java.time.LocalDateTime;
6
import java.time.ZoneId;
7
import java.time.format.DateTimeFormatter;
8
import java.util.Collection;
9
import java.util.Date;
10
import java.util.Map;
11
import java.util.zip.ZipEntry;
12
import java.util.zip.ZipOutputStream;
13

    
14
import com.google.common.base.Function;
15
import com.google.common.collect.Iterables;
16
import com.google.common.collect.Lists;
17
import com.google.common.collect.Maps;
18
import com.google.common.collect.Multimap;
19
import com.mongodb.BasicDBObject;
20
import com.mongodb.DBObject;
21
import com.mongodb.MongoClient;
22
import com.mongodb.WriteConcern;
23
import com.mongodb.client.MongoCollection;
24
import com.mongodb.client.MongoDatabase;
25
import eu.dnetlib.data.mapreduce.JobParams;
26
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
import eu.dnetlib.data.mapreduce.hbase.oai.utils.RecordFieldsExtractor;
32
import eu.dnetlib.miscutils.functional.xml.IndentXmlString;
33
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

    
40
public class OaiFeedMapper extends Mapper<Text, Text, NullWritable, NullWritable> {
41

    
42
	enum RecordStatus {
43
		NEW, UPDATED, UNCHANGED;
44
	}
45

    
46
	private MongoCollection<DBObject> collection;
47
	private MongoCollection<DBObject> discardedCollection;
48
	private OAIConfigurationStringReader oaiConfigurationReader;
49
	private OAIConfiguration oaiConfiguration;
50

    
51
	private Date feedDate;
52

    
53
	private MongoSetCollection mongoSetCollection;
54

    
55
	private RecordFieldsExtractor extractor;
56

    
57
	// these are set in the setup
58
	private String format;
59
	private String interpretation;
60
	private String layout;
61
	private Map<String, PublisherField> fieldsToIndex = Maps.newHashMap();
62

    
63
	private String duplicateXPath;
64
	private boolean skipDuplicates;
65

    
66
	private MongoClient mongo;
67

    
68
	private Collection<String> enrichmentXPaths;
69

    
70
	@Override
71
	protected void setup(final Context context) throws UnknownHostException {
72

    
73
		String host = context.getConfiguration().get("services.publisher.oai.host");
74
		String port = context.getConfiguration().get("services.publisher.oai.port");
75
		String db = context.getConfiguration().get("services.publisher.oai.db");
76
		String collectionName = context.getConfiguration().get("services.publisher.oai.collection");
77

    
78
		System.out.println("Mongodb client params");
79
		System.out.println("host: " + host);
80
		System.out.println("port: " + port);
81
		System.out.println("db: " + db);
82
		System.out.println("collection: " + collectionName);
83

    
84
		String[] formatLayoutInterp = collectionName.split("-");
85
		format = formatLayoutInterp[0];
86
		layout = formatLayoutInterp[1];
87
		interpretation = formatLayoutInterp[2];
88

    
89
		String oaiConfigurationProfile = context.getConfiguration().get("oaiConfiguration");
90
		System.out.println("oaiConfiguration:\n" + IndentXmlString.apply(oaiConfigurationProfile));
91
		oaiConfigurationReader = new OAIConfigurationStringReader(oaiConfigurationProfile);
92
		oaiConfiguration = oaiConfigurationReader.getOaiConfiguration();
93

    
94
		System.out.println("parsed configuration:" + oaiConfiguration.toString());
95

    
96
		mongo = new MongoClient(host, Integer.parseInt(port));
97
		MongoDatabase mongoDB = mongo.getDatabase(db);
98
		//DB mongoDB = mongo.getDB(db);
99
		collection = mongoDB.getCollection(collectionName, DBObject.class).withWriteConcern(WriteConcern.UNACKNOWLEDGED);
100
		discardedCollection = mongoDB.getCollection("discarded-" + collectionName, DBObject.class).withWriteConcern(WriteConcern.UNACKNOWLEDGED);
101
		mongoSetCollection = new MongoSetCollection(mongo);
102

    
103
		duplicateXPath = context.getConfiguration().get("services.publisher.oai.duplicateXPath");
104
		skipDuplicates = Boolean.parseBoolean(context.getConfiguration().get("services.publisher.oai.skipDuplicates"));
105

    
106
		enrichmentXPaths = oaiConfiguration.getEnrichmentXPathsFor(format, layout, interpretation);
107
		Collection<PublisherField> indexFields = oaiConfiguration.getFieldsFor(format, layout, interpretation);
108
		extractor = new RecordFieldsExtractor(Lists.newArrayList(indexFields));
109
		extractor.setDuplicateXPath(duplicateXPath);
110
		extractor.setSkipDuplicates(skipDuplicates);
111

    
112
		for (PublisherField field : indexFields) {
113
			fieldsToIndex.put(field.getFieldName(), field);
114
		}
115

    
116
		String feedDateString = context.getConfiguration().get(JobParams.OAI_FEED_DATE);
117
		feedDate = parseDate(feedDateString);
118
	}
119

    
120
	@Override
121
	protected void map(final Text key, final Text value, final Context context) throws IOException, InterruptedException {
122
		String recordKey = key.toString();
123
		String recordBody = value.toString();
124
		if (StringUtils.isBlank(recordBody)) {
125
			discard(context, recordKey, recordBody, "blank body");
126
		} else {
127
			Multimap<String, String> recordFields = extractor.extractFields(recordBody, enrichmentXPaths);
128
			String id;
129
			String oaiID;
130
			if (checkRecordFields(recordFields, context, recordKey, recordBody)) {
131
				id = recordFields.get(OAIConfigurationReader.ID_FIELD).iterator().next();
132
				oaiID = getOAIIdentifier(id);
133
				handleRecord(context, oaiID, recordBody, recordFields);
134
			}
135
		}
136
	}
137

    
138
	public boolean checkRecordFields(final Multimap<String, String> recordFields, final Context context, final String recordKey, final String recordBody) {
139
		if ((recordFields == null)) {
140
			context.getCounter("oai", "invalid").increment(1);
141
			return false;
142
		}
143
		if (recordFields.containsEntry("duplicate", "true")) {
144
			if (skipDuplicates) {
145
				context.getCounter("oai", "discardedDuplicate").increment(1);
146
				return false;
147
			} else return true;
148
		}
149
		if (!recordFields.containsKey(OAIConfigurationReader.ID_FIELD)) {
150
			discard(context, recordKey, recordBody, "missing " + OAIConfigurationReader.ID_FIELD);
151
			return false;
152
		}
153
		return true;
154
	}
155

    
156
	private void handleRecord(final Context context, final String oaiID, final String record, final Multimap<String, String> recordProperties) {
157
		DBObject obj = this.createBasicObject(oaiID, record, recordProperties,context);
158
		if (obj != null) { // it can be null if the compression did not succeeded: counter is updated in the compress method in that case
159
			//let's use the date of collection and transformation for those records that have them.
160
			//for the rest, we'll keep the feedDate...
161
			Collection<String> collectionDates = recordProperties.get("dateOfCollection");
162
			Collection<String> transDates = recordProperties.get("dateOfTransformation");
163
			Date collDate = feedDate;
164
			if(collectionDates != null && !collectionDates.isEmpty()){
165
				String collDateString = collectionDates.iterator().next();
166
				if(StringUtils.isNotBlank(collDateString))
167
					collDate = parseDate(collDateString);
168
			}
169
			obj.put(OAIConfigurationReader.LAST_COLLECTION_DATE_FIELD, collDate);
170
			Date timestamp = feedDate;
171
			if(transDates != null && !transDates.isEmpty()){
172
				String transDateString = transDates.iterator().next();
173
				if(StringUtils.isNotBlank(transDateString))
174
					timestamp = parseDate(transDateString);
175
			}
176
			obj.put(OAIConfigurationReader.DATESTAMP_FIELD, timestamp);
177
			//the updated_field must stay false, as the oai store is empty.
178
			obj.put(OAIConfigurationReader.UPDATED_FIELD, false);
179
			collection.insertOne(obj);
180
			context.getCounter("oai", "total").increment(1);
181
		}
182
	}
183

    
184
	protected Date parseDate(final String date) {
185
		// 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
186
		try {
187
			LocalDateTime d = LocalDateTime.parse(date, DateTimeFormatter.ISO_ZONED_DATE_TIME);
188
			return Date.from(d.atZone(ZoneId.systemDefault()).toInstant());
189
		}catch(Exception dateException){
190
			try {
191
				return org.apache.commons.lang.time.DateUtils.parseDate(
192
						date,
193
						new String[] { "yyyy-MM-dd", "yyyy-MM-dd'T'HH:mm:ssXXX", "yyyy-MM-dd'T'HH:mm:ss.SSSX", "yyyy-MM-dd'T'HH:mm:ssZ",
194
								"yyyy-MM-dd'T'HH:mm:ss.SX" });
195
			}catch(Exception dateException2){
196
				dateException2.printStackTrace(System.err);
197
				throw new RuntimeException(dateException2);
198
			}
199
		}
200
	}
201

    
202
	private void discard(final Context context, final String recordKey, final String recordBody, final String reason) {
203
		context.getCounter("oai", reason).increment(1);
204
		discardedCollection.insertOne(new BasicDBObject("id", recordKey).append(OAIConfigurationReader.BODY_FIELD, recordBody));
205
	}
206

    
207
	private String getOAIIdentifier(final String id) {
208
		return oaiConfiguration.getIdScheme() + ":" + oaiConfiguration.getIdNamespace() + ":" + id;
209
	}
210

    
211
	protected DBObject createBasicObject(final String oaiID, final String record, final Multimap<String, String> recordProperties, final Context context) {
212
		DBObject obj = new BasicDBObject();
213
		for (final String key : recordProperties.keySet()) {
214
			if (key.equals(OAIConfigurationReader.ID_FIELD)) {
215
				obj.put(key, oaiID);
216
			} else {
217
				Collection<String> values = recordProperties.get(key);
218
				if (key.equals(OAIConfigurationReader.SET_FIELD)) {
219

    
220
					Iterable<String> setSpecs = Iterables.transform(values, new Function<String, String>() {
221

    
222
						@Override
223
						public String apply(final String s) {
224
							return mongoSetCollection.normalizeSetSpec(s);
225
						}
226

    
227
					});
228
					obj.put(key, setSpecs);
229
				} else {
230
					PublisherField keyField = fieldsToIndex.get(key);
231
					if (keyField == null) {
232
						context.getCounter("oai", key + " found for record but not in configuration. Assuming it is repeatable.").increment(1);
233
					}
234
					// let's check if the key is the name of a repeatable field or not
235
					if ((keyField != null) && !keyField.isRepeatable()) {
236
						if ((values != null) && !values.isEmpty()) {
237
							obj.put(key, values.iterator().next());
238
						}
239
					} else {
240
						obj.put(key, values);
241
					}
242
				}
243
			}
244
		}
245

    
246
		Binary compressedRecordBody = createCompressRecord(context, oaiID, record);
247
		if (compressedRecordBody != null) {
248
			obj.put(OAIConfigurationReader.BODY_FIELD, compressedRecordBody);
249
			obj.put(OAIConfigurationReader.DELETED_FIELD, false);
250
			return obj;
251
		} else return null;
252
	}
253

    
254
	public Binary createCompressRecord(final Context context, final String recordKey, final String recordBody) {
255
		try {
256
			ByteArrayOutputStream os = new ByteArrayOutputStream();
257
			ZipOutputStream zos = new ZipOutputStream(os);
258
			ZipEntry entry = new ZipEntry(OAIConfigurationReader.BODY_FIELD);
259
			zos.putNextEntry(entry);
260
			zos.write(recordBody.getBytes());
261
			zos.closeEntry();
262
			zos.flush();
263
			zos.close();
264
			return new Binary(os.toByteArray());
265
		} catch (IOException e) {
266
			discard(context, recordKey, recordBody, "cannot compress");
267
			return null;
268
		}
269
	}
270

    
271
	@Override
272
	protected void cleanup(final Context context) throws IOException, InterruptedException {
273

    
274
		super.cleanup(context);
275
	}
276

    
277
	public MongoCollection<DBObject> getCollection() {
278
		return collection;
279
	}
280

    
281
	public void setCollection(final MongoCollection<DBObject> collection) {
282
		this.collection = collection;
283
	}
284

    
285
	public MongoCollection<DBObject> getDiscardedCollection() {
286
		return discardedCollection;
287
	}
288

    
289
	public void setDiscardedCollection(final MongoCollection<DBObject> discardedCollection) {
290
		this.discardedCollection = discardedCollection;
291
	}
292

    
293
	public OAIConfigurationStringReader getOaiConfigurationReader() {
294
		return oaiConfigurationReader;
295
	}
296

    
297
	public void setOaiConfigurationReader(final OAIConfigurationStringReader oaiConfigurationReader) {
298
		this.oaiConfigurationReader = oaiConfigurationReader;
299
	}
300

    
301
	public OAIConfiguration getOaiConfiguration() {
302
		return oaiConfiguration;
303
	}
304

    
305
	public void setOaiConfiguration(final OAIConfiguration oaiConfiguration) {
306
		this.oaiConfiguration = oaiConfiguration;
307
	}
308

    
309
	public Date getFeedDate() {
310
		return feedDate;
311
	}
312

    
313
	public void setFeedDate(final Date feedDate) {
314
		this.feedDate = feedDate;
315
	}
316

    
317
	public MongoSetCollection getMongoSetCollection() {
318
		return mongoSetCollection;
319
	}
320

    
321
	public void setMongoSetCollection(final MongoSetCollection mongoSetCollection) {
322
		this.mongoSetCollection = mongoSetCollection;
323
	}
324

    
325
	public String getDuplicateXPath() {
326
		return duplicateXPath;
327
	}
328

    
329
	public void setDuplicateXPath(final String duplicateXPath) {
330
		this.duplicateXPath = duplicateXPath;
331
	}
332

    
333
	public boolean isSkipDuplicates() {
334
		return skipDuplicates;
335
	}
336

    
337
	public void setSkipDuplicates(final boolean skipDuplicates) {
338
		this.skipDuplicates = skipDuplicates;
339
	}
340

    
341
}
    (1-1/1)