1
|
package eu.dnetlib.data.mdstore.modular.mongodb;
|
2
|
|
3
|
import java.util.List;
|
4
|
import java.util.Map;
|
5
|
import javax.annotation.Nullable;
|
6
|
|
7
|
import com.google.common.base.Predicate;
|
8
|
import com.google.common.collect.Lists;
|
9
|
import com.google.common.collect.Maps;
|
10
|
import com.mongodb.BasicDBObject;
|
11
|
import com.mongodb.DBObject;
|
12
|
import com.mongodb.WriteConcern;
|
13
|
import com.mongodb.client.MongoCollection;
|
14
|
import com.mongodb.client.model.*;
|
15
|
import eu.dnetlib.data.mdstore.MDStoreServiceException;
|
16
|
import eu.dnetlib.data.mdstore.modular.MDFormatDescription;
|
17
|
import eu.dnetlib.data.mdstore.modular.RecordParser;
|
18
|
import eu.dnetlib.data.mdstore.modular.mongodb.utils.IndexFieldRecordParser;
|
19
|
import eu.dnetlib.data.mdstore.modular.mongodb.utils.IndexFieldRecordParserException;
|
20
|
import org.apache.commons.logging.Log;
|
21
|
import org.apache.commons.logging.LogFactory;
|
22
|
|
23
|
import static eu.dnetlib.data.mdstore.modular.MDStoreConstants.*;
|
24
|
|
25
|
public class MongoBulkWritesManager {
|
26
|
|
27
|
private static final Log log = LogFactory.getLog(MongoBulkWritesManager.class);
|
28
|
private final boolean discardRecords;
|
29
|
private final boolean indexRecords;
|
30
|
private final IndexFieldRecordParser indexFieldRecordParser = new IndexFieldRecordParser();
|
31
|
private final List<MDFormatDescription> mdref;
|
32
|
private RecordParser recordParser;
|
33
|
private MongoCollection<DBObject> validCollection;
|
34
|
private List<WriteModel<DBObject>> validBulkOperationList;
|
35
|
private BulkWriteOptions writeOptions;
|
36
|
private MongoCollection<DBObject> discardedCollection;
|
37
|
private List<WriteModel<DBObject>> discardedBulkOperationList;
|
38
|
private int bulkSize;
|
39
|
|
40
|
public MongoBulkWritesManager(final MongoCollection<DBObject> collection,
|
41
|
final MongoCollection<DBObject> discardedCollection,
|
42
|
final List<MDFormatDescription> mdref,
|
43
|
final int bulkSize,
|
44
|
final RecordParser parser,
|
45
|
final boolean discardRecords) {
|
46
|
this.validCollection = collection.withWriteConcern(WriteConcern.ACKNOWLEDGED);
|
47
|
this.validBulkOperationList = Lists.newArrayList();
|
48
|
|
49
|
this.discardedCollection = discardedCollection.withWriteConcern(WriteConcern.ACKNOWLEDGED);
|
50
|
this.discardedBulkOperationList = Lists.newArrayList();
|
51
|
|
52
|
this.bulkSize = bulkSize;
|
53
|
this.recordParser = parser;
|
54
|
this.discardRecords = discardRecords;
|
55
|
this.mdref = mdref;
|
56
|
|
57
|
this.indexRecords = (this.mdref != null && !this.mdref.isEmpty());
|
58
|
this.writeOptions = new BulkWriteOptions().ordered(false);
|
59
|
}
|
60
|
|
61
|
public void insert(final String record) throws MDStoreServiceException {
|
62
|
Map<String, String> recordProperties = null;
|
63
|
try {
|
64
|
recordProperties = recordParser.parseRecord(record);
|
65
|
|
66
|
} catch (Throwable e) {
|
67
|
if (discardRecords) {
|
68
|
log.debug("unhandled exception: " + e.getMessage());
|
69
|
discardRecord(record);
|
70
|
}
|
71
|
}
|
72
|
Map<String, List<String>> indexRecordField = null;
|
73
|
try {
|
74
|
if (indexRecords) {
|
75
|
indexRecordField = indexFieldRecordParser.parseRecord(record, mdref);
|
76
|
}
|
77
|
} catch (IndexFieldRecordParserException e) {
|
78
|
// could not index record fields
|
79
|
throw new MDStoreServiceException("Are you using the correct type of store / index definition for the records in " + validCollection.getNamespace() + " ?", e);
|
80
|
}
|
81
|
|
82
|
log.debug("found props: " + recordProperties);
|
83
|
if (recordProperties.containsKey(ID)) {
|
84
|
final DBObject obj = buildDBObject(record, recordProperties, indexRecordField);
|
85
|
if (log.isDebugEnabled()) {
|
86
|
log.debug("Saving object" + obj);
|
87
|
}
|
88
|
validBulkOperationList.add(new ReplaceOneModel(new BasicDBObject(ID, obj.get(ID)), obj, new UpdateOptions().upsert(true)));
|
89
|
if (((validBulkOperationList.size() % bulkSize) == 0) && !validBulkOperationList.isEmpty()) {
|
90
|
validCollection.bulkWrite(validBulkOperationList, writeOptions);
|
91
|
validBulkOperationList.clear();
|
92
|
}
|
93
|
} else {
|
94
|
if (discardRecords) {
|
95
|
log.debug("parsed record seems invalid");
|
96
|
discardRecord(record);
|
97
|
}
|
98
|
}
|
99
|
}
|
100
|
|
101
|
private void discardRecord(final String record) {
|
102
|
discardedBulkOperationList.add(new InsertOneModel(new BasicDBObject(BODY, record)));
|
103
|
|
104
|
if (((discardedBulkOperationList.size() % bulkSize) == 0) && !discardedBulkOperationList.isEmpty()) {
|
105
|
discardedCollection.bulkWrite(discardedBulkOperationList, writeOptions);
|
106
|
discardedBulkOperationList.clear();
|
107
|
}
|
108
|
}
|
109
|
|
110
|
public void flushBulks() {
|
111
|
//setting to journaled write concern to be sure that when the write returns everything has been flushed to disk (https://docs.mongodb.org/manual/faq/developers/#when-does-mongodb-write-updates-to-disk)
|
112
|
//the explicit fsync command can't be run anymore: 'Command failed with error 13: 'fsync may only be run against the admin database.'
|
113
|
if (!validBulkOperationList.isEmpty()) {
|
114
|
validCollection = getCollectionWithWriteConcern(validCollection, WriteConcern.JOURNALED);
|
115
|
validCollection.bulkWrite(validBulkOperationList, writeOptions);
|
116
|
}
|
117
|
if (!discardedBulkOperationList.isEmpty()) {
|
118
|
discardedCollection = getCollectionWithWriteConcern(discardedCollection, WriteConcern.JOURNALED);
|
119
|
discardedCollection.bulkWrite(discardedBulkOperationList, writeOptions);
|
120
|
}
|
121
|
//setting write concern back to ACKNOWLEDGE to avoid the execution of future writes all in Journaled mode
|
122
|
validCollection = getCollectionWithWriteConcern(validCollection, WriteConcern.ACKNOWLEDGED);
|
123
|
discardedCollection = getCollectionWithWriteConcern(discardedCollection, WriteConcern.ACKNOWLEDGED);
|
124
|
}
|
125
|
|
126
|
protected DBObject buildDBObject(final String record, final Map<String, String> recordProperties, final Map<String, List<String>> indexFieldProperties) {
|
127
|
final DBObject obj = new BasicDBObject();
|
128
|
obj.put(ID, recordProperties.get(ID));
|
129
|
obj.put(ORIGINALID, recordProperties.get(ORIGINALID));
|
130
|
obj.put(BODY, record);
|
131
|
obj.put(TIMESTAMP, Long.valueOf(recordProperties.get(TIMESTAMP)));
|
132
|
if (indexFieldProperties != null)
|
133
|
obj.putAll(Maps.filterKeys(indexFieldProperties, new Predicate<String>() {
|
134
|
//ensure we do not override the mandatory fields above with some unexpected value
|
135
|
@Override
|
136
|
public boolean apply(@Nullable final String s) {
|
137
|
return !s.equalsIgnoreCase(ID) && !s.equalsIgnoreCase(ORIGINALID) && !s.equalsIgnoreCase(BODY) && !s.equalsIgnoreCase(TIMESTAMP);
|
138
|
}
|
139
|
}));
|
140
|
return obj;
|
141
|
}
|
142
|
|
143
|
private MongoCollection<DBObject> getCollectionWithWriteConcern(MongoCollection<DBObject> collection, WriteConcern writeConcern) {
|
144
|
return collection.withWriteConcern(writeConcern);
|
145
|
}
|
146
|
|
147
|
}
|