1
|
package eu.dnetlib.data.mapreduce.dedup;
|
2
|
|
3
|
import java.util.LinkedList;
|
4
|
import java.util.List;
|
5
|
import java.util.Map;
|
6
|
import java.util.PriorityQueue;
|
7
|
import java.util.Queue;
|
8
|
import java.util.UUID;
|
9
|
|
10
|
import org.apache.commons.io.IOUtils;
|
11
|
import org.junit.Before;
|
12
|
import org.junit.Test;
|
13
|
import org.springframework.core.io.ClassPathResource;
|
14
|
|
15
|
import com.google.common.collect.Lists;
|
16
|
import com.google.common.collect.Maps;
|
17
|
|
18
|
import eu.dnetlib.pace.clustering.NGramUtils;
|
19
|
import eu.dnetlib.pace.config.Type;
|
20
|
import eu.dnetlib.pace.model.Field;
|
21
|
import eu.dnetlib.pace.model.MapDocument;
|
22
|
import eu.dnetlib.pace.model.MapDocumentComparator;
|
23
|
|
24
|
public class TitleOrderingTest {
|
25
|
|
26
|
private List<MapDocument> results = Lists.newArrayList();
|
27
|
|
28
|
@Before
|
29
|
public void setUp() throws Exception {
|
30
|
|
31
|
@SuppressWarnings("unchecked")
|
32
|
final List<String> lines = IOUtils.readLines(new ClassPathResource("eu/dnetlib/data/mapreduce/dedup/titles.txt").getInputStream());
|
33
|
for (String title : lines) {
|
34
|
Map<String, List<Field>> fieldMap = Maps.newHashMap();
|
35
|
List<Field> list = Lists.newArrayList();
|
36
|
list.add(new Field(Type.String, "title", title));
|
37
|
fieldMap.put("title", list);
|
38
|
results.add(new MapDocument("id-" + UUID.randomUUID(), fieldMap));
|
39
|
}
|
40
|
}
|
41
|
|
42
|
@Test
|
43
|
public void test() {
|
44
|
|
45
|
final Queue<MapDocument> queue = new PriorityQueue<MapDocument>(100, new MapDocumentComparator("title"));
|
46
|
|
47
|
queue.addAll(results);
|
48
|
|
49
|
Queue<MapDocument> queue2 = simplifyQueue(queue);
|
50
|
|
51
|
while (!queue2.isEmpty()) {
|
52
|
MapDocument doc = queue2.remove();
|
53
|
System.out.println(doc.values("title").get(0).getValue());
|
54
|
}
|
55
|
}
|
56
|
|
57
|
private Queue<MapDocument> simplifyQueue(final Queue<MapDocument> queue) {
|
58
|
final Queue<MapDocument> q = new LinkedList<MapDocument>();
|
59
|
|
60
|
String fieldRef = "";
|
61
|
List<MapDocument> tempResults = Lists.newArrayList();
|
62
|
|
63
|
while (!queue.isEmpty()) {
|
64
|
MapDocument result = queue.remove();
|
65
|
|
66
|
if (!result.values("title").isEmpty()) {
|
67
|
String field = NGramUtils.cleanupForOrdering(result.values("title").get(0).getValue().toString());
|
68
|
if (field.equals(fieldRef)) {
|
69
|
tempResults.add(result);
|
70
|
} else {
|
71
|
if (tempResults.size() < 5) {
|
72
|
q.addAll(tempResults);
|
73
|
} else {
|
74
|
System.out.println("Skipped field: " + fieldRef + " - size: " + tempResults.size());
|
75
|
}
|
76
|
tempResults.clear();
|
77
|
tempResults.add(result);
|
78
|
fieldRef = field;
|
79
|
}
|
80
|
}
|
81
|
}
|
82
|
if (tempResults.size() < 5) {
|
83
|
q.addAll(tempResults);
|
84
|
} else {
|
85
|
System.out.println("Skipped field: " + fieldRef + " - size: " + tempResults.size());
|
86
|
}
|
87
|
|
88
|
return q;
|
89
|
}
|
90
|
|
91
|
}
|