Lucene In Action ch 6(I)
笔记 --自定义排序
----- 2006-2-16
使用
Lucene
来搜索内容,搜索结果的显示顺序当然是比较重要的.Lucene中Build-in的几个排序定义在大多数情况下是不适合我们使用的.要适合自己的应用程序的场景,就只能自定义排序功能,本节我们就来看看在Lucene中如何实现自定义排序功能.
Lucene
中的自定义排序功能和Java集合中的自定义排序的实现方法差不多,都要实现一下比较接口. 在Java中只要实现Comparable接口就可以了.但是在Lucene中要实现SortComparatorSource接口和ScoreDocComparator接口.在了解具体实现方法之前先来看看这两个接口的定义吧.
SortComparatorSource
接口的功能是返回一个用来排序ScoreDocs的comparator(Expert: returns a comparator for sorting ScoreDocs).该接口只定义了一个方法.如下:
public ScoreDocComparatornewComparator(IndexReader reader,String fieldname) throws IOException
Creates a comparator for the field in the given index.
Parameters:
reader - Index to create comparator for.
fieldname - Field to create comparator for.
Returns:
Comparator of ScoreDoc objects.
Throws:
IOException - If an error occurs reading the index.
该方法
只是创造一个ScoreDocComparator实例用来实现排序.所以我们还要实现ScoreDocComparator接口.来看看ScoreDocComparator接口.功能是比较来两个ScoreDoc 对象来排序(Compares two ScoreDoc objects for sorting) 里面定义了两个Lucene实现的静态实例.如下:
public static final ScoreDocComparatorRELEVANCE
Special comparator for sorting hits according to computed relevance (document score).
public static final ScoreDocComparatorINDEXORDER
Special comparator for sorting hits according to index order (document number).
有3个方法与排序相关,需要我们实现 分别如下:
public int compare(ScoreDoc i,ScoreDoc j)
Compares two ScoreDoc objects and returns a result indicating their sort order.
Parameters:
i - First ScoreDoc
j - Second ScoreDoc
Returns:
-1 if i should come before j
1 if i should come after j
0 if they are equal
public ComparablesortValue(ScoreDoc i)
Returns the value used to sort the given document. The object returned must implement the java.io.Serializable interface. This is used by multisearchers to determine how to collate results from their searchers.
Parameters:
i - Document
Returns:
Serializable object
public int sortType()
Returns the type of sort. Should return SortField.SCORE, SortField.DOC, SortField.STRING, SortField.INTEGER, SortField.FLOAT or SortField.CUSTOM. It is not valid to return SortField.AUTO. This is used by multisearchers to determine how to collate results from their searchers.
Returns:
One of the constants in SortField.
看个例子吧!
该例子为Lucene in Action中的一个实现,用来搜索距你最近的餐馆的名字. 餐馆坐标用字符串"x,y"来存储.如下图:
Figure 6.1 Which Mexican restaurant is closest to home (at 0,0) or work (at 10,10)?
此中情况下 Lucene中Build-in Sorting 实现就不可行了,看看如何自己实现吧.
01
package lia.extsearch.sorting;
02
03
import org.apache.lucene.search.SortComparatorSource;
04
import org.apache.lucene.search.ScoreDoc;
05
import org.apache.lucene.search.SortField;
06
import org.apache.lucene.search.ScoreDocComparator;
07
import org.apache.lucene.index.IndexReader;
08
import org.apache.lucene.index.TermEnum;
09
import org.apache.lucene.index.Term;
10
import org.apache.lucene.index.TermDocs;
11
12
import java.io.IOException;
13
// DistanceComparatorSource
实现了
SortComparatorSource
接口
14
public class DistanceComparatorSource implements SortComparatorSource {
15
// x y
用来保存
坐标位置
16
private int x;
17
private int y;
18
19
public DistanceComparatorSource(int x, int y) {
20
this.x = x;
21
this.y = y;
22
}
23
//
返回
ScoreDocComparator
用来实现排序功能
24
public ScoreDocComparator newComparator(IndexReader reader, String fieldname)
25
throws IOException {
26
return new DistanceScoreDocLookupComparator(reader, fieldname, x, y);
27
}
28
29
//DistanceScoreDocLookupComparator
实现了
ScoreDocComparator
用来排序
30
private static class DistanceScoreDocLookupComparator implements
31
ScoreDocComparator {
32
private float[] distances; //
保存每个餐馆到指定点的距离
33
34
//
构造函数
,
构造函数在这里几乎完成所有的准备工作
.
35
public DistanceScoreDocLookupComparator(IndexReader reader,
36
String fieldname, int x, int y) throws IOException {
37
38
final TermEnum enumerator = reader.terms(new Term(fieldname, ""));
39
distances = new float[reader.maxDoc()]; //
初始化
distances
40
if (distances.length > 0) {
41
TermDocs termDocs = reader.termDocs();
42
try {
43
if (enumerator.term() == null) {
44
throw new RuntimeException("no terms in field "
45
+ fieldname);
46
}
47
int i = 0,j = 0;
48
do {
49
System.out.println("in do-while :" + i ++);
50
51
Term term = enumerator.term(); //
取出每一个
Term
52
if (term.field() != fieldname) //
与给定的域不符合则比较下一个
53
break;
54
//Sets this to the data for the current term in a TermEnum.
55
//This may be optimized in some implementations.
56
termDocs.seek(enumerator); //
参考
TermDocs Doc
57
while (termDocs.next()) {
58
System.out.println(" in while :" + j ++);
59
System.out.println(" in while ,Term :" + term.toString());
60
61
String[] xy = term.text().split(","); //
去处
x y
62
int deltax = Integer.parseInt(xy[0]) - x;
63
int deltay = Integer.parseInt(xy[1]) - y;
64
//
计算距离
65
distances[termDocs.doc()] = (float) Math
66
.sqrt(deltax * deltax + deltay * deltay);
67
}
68
} while (enumerator.next());
69
} finally {
70
termDocs.close();
71
}
72
}
73
}
74
75
//
有上面的构造函数的准备
这里就比较简单了
76
public int compare(ScoreDoc i, ScoreDoc j) {
77
if (distances[i.doc] < distances[j.doc])
78
return -1;
79
if (distances[i.doc] > distances[j.doc])
80
return 1;
81
return 0;
82
}
83
84
//
返回距离
85
public Comparable sortValue(ScoreDoc i) {
86
return new Float(distances[i.doc]);
87
}
88
89
//
指定
SortType
90
public int sortType() {
91
return SortField.FLOAT;
92
}
93
}
94
95
public String toString() {
96
return "Distance from (" + x + "," + y + ")";
97
}
98
99
}
这是一个实现了上面两个接口的两个类, 里面带有详细注释, 可以看出 自定义排序并不是很难的. 该实现能否正确实现,我们来看看测试代码能否通过吧.
001
package lia.extsearch.sorting;
002
003
import junit.framework.TestCase;
004
import org.apache.lucene.analysis.WhitespaceAnalyzer;
005
import org.apache.lucene.document.Document;
006
import org.apache.lucene.document.Field;
007
import org.apache.lucene.index.IndexWriter;
008
import org.apache.lucene.index.Term;
009
import org.apache.lucene.search.FieldDoc;
010
import org.apache.lucene.search.Hits;
011
import org.apache.lucene.search.IndexSearcher;
012
import org.apache.lucene.search.Query;
013
import org.apache.lucene.search.ScoreDoc;
014
import org.apache.lucene.search.Sort;
015
import org.apache.lucene.search.SortField;
016
import org.apache.lucene.search.TermQuery;
017
import org.apache.lucene.search.TopFieldDocs;
018
import org.apache.lucene.store.RAMDirectory;
019
020
import java.io.IOException;
021
022
import lia.extsearch.sorting.DistanceComparatorSource;
023
//
测试
自定义排序的实现
024
public class DistanceSortingTest extends TestCase {
025
private RAMDirectory directory;
026
027
private IndexSearcher searcher;
028
029
private Query query;
030
031
//
建立测试环境
032
protected void setUp() throws Exception {
033
directory = new RAMDirectory();
034
IndexWriter writer = new IndexWriter(directory,
035
new WhitespaceAnalyzer(), true);
036
addPoint(writer, "El Charro", "restaurant", 1, 2);
037
addPoint(writer, "Cafe Poca Cosa", "restaurant", 5, 9);
038
addPoint(writer, "Los Betos", "restaurant", 9, 6);
039
addPoint(writer, "Nico's Taco Shop", "restaurant", 3, 8);
040
041
writer.close();
042
043
searcher = new IndexSearcher(directory);
044
045
query = new TermQuery(new Term("type", "restaurant"));
046
}
047
048
private void addPoint(IndexWriter writer, String name, String type, int x,
049
int y) throws IOException {
050
Document doc = new Document();
051
doc.add(Field.Keyword("name", name));
052
doc.add(Field.Keyword("type", type));
053
doc.add(Field.Keyword("location", x + "," + y));
054
writer.addDocument(doc);
055
}
056
057
public void testNearestRestaurantToHome() throws Exception {
058
//
使用
DistanceComparatorSource
来构造一个
SortField
059
Sort sort = new Sort(new SortField("location",
060
new DistanceComparatorSource(0, 0)));
061
062
Hits hits = searcher.search(query, sort); //
搜索
063
064
//
测试
065
assertEquals("closest", "El Charro", hits.doc(0).get("name"));
066
assertEquals("furthest", "Los Betos", hits.doc(3).get("name"));
067
}
068
069
public void testNeareastRestaurantToWork() throws Exception {
070
Sort sort = new Sort(new SortField("location",
071
new DistanceComparatorSource(10, 10))); //
工作的坐标
10,10
072
073
//
上面的测试实现了自定义排序
,
但是并不能访问自定义排序的更详细信息
,
利用
074
//TopFieldDocs
可以进一步访问相关信息
075
TopFieldDocs docs = searcher.search(query, null, 3, sort);
076
077
assertEquals(4, docs.totalHits);
078
assertEquals(3, docs.scoreDocs.length);
079
080
//
取得
FieldDoc
利用
FieldDoc
可以取得关于排序的更详细信息
请查看
FieldDoc Doc
081
FieldDoc fieldDoc = (FieldDoc) docs.scoreDocs[0];
082
083
assertEquals("(10,10) -> (9,6) = sqrt(17)", new Float(Math.sqrt(17)),
084
fieldDoc.fields[0]);
085
086
Document document = searcher.doc(fieldDoc.doc);
087
assertEquals("Los Betos", document.get("name"));
088
089
dumpDocs(sort, docs); //
显示相关信息
090
}
091
092
//
显示有关排序的信息
093
private void dumpDocs(Sort sort, TopFieldDocs docs) throws IOException {
094
System.out.println("Sorted by: " + sort);
095
ScoreDoc[] scoreDocs = docs.scoreDocs;
096
for (int i = 0; i < scoreDocs.length; i++) {
097
FieldDoc fieldDoc = (FieldDoc) scoreDocs[i];
098
Float distance = (Float) fieldDoc.fields[0];
099
Document doc = searcher.doc(fieldDoc.doc);
100
System.out.println(" " + doc.get("name") + " @ ("
101
+ doc.get("location") + ") -> " + distance);
102
}
103
}
104
}
完全通过测试,
输入信息如下:想进一步了解详细信息的可以研究一下:
in do-while :0
in while :0
in while ,Term :location:1,2
in do-while :1
in while :1
in while ,Term :location:3,8
in do-while :2
in while :2
in while ,Term :location:5,9
in do-while :3
in while :3
in while ,Term :location:9,6
in do-while :4
in do-while :0
in while :0
in while ,Term :location:1,2
in do-while :1
in while :1
in while ,Term :location:3,8
in do-while :2
in while :2
in while ,Term :location:5,9
in do-while :3
in while :3
in while ,Term :location:9,6
in do-while :4
Sorted by: <custom:"location": Distance from (10,10)>
Los Betos @ (9,6) -> 4.1231055
Cafe Poca Cosa @ (5,9) -> 5.0990195
Nico's Taco Shop @ (3,8) -> 7.28011
如果要想取得测试的详细参考信息可以参考
testNeareastRestaurantToWork
方法的实现
.
有上面可以看出要自定义实现排序并不是很难的.
下面来看看HitCollector.
一般情况下搜索结果只显示最重要的一些结果,但有时用户可能想显示所有匹配的搜索结果而不访问其内容.这中情况下使用自定义的HitCollector是高效的实现.
下面来看看一个测试例子.在该例子中我们实现了
BookLinkCollector
一个自定义的HitCollector,里面有一个Map 保存了符合查询条件的 URL 和相应的booktitle ,HitCollector中有个函数要实现 collect:其doc如下:
public abstract void collect(int doc, float score)
Called once for every non-zero scoring document, with the document number and its score.
If, for example, an application wished to collect all of the hits for a query in a BitSet, then it might:
Searcher searcher = new IndexSearcher(indexReader);
final BitSet bits = new BitSet(indexReader.maxDoc());
searcher.search(query, new HitCollector() {
public void collect(int doc, float score) {
bits.set(doc);
}
});
Note: This is called in an inner search loop. For good search performance, implementations of this method should not call Searchable.doc(int) or IndexReader.document(int) on every document number encountered. Doing so can slow searches by an order of magnitude or more.
Note: The score passed to this method is a raw score. In other words, the score will not necessarily be a float whose value is between 0 and 1.
下面来看看BookLinkCollector的实现:
01
package lia.extsearch.hitcollector;
02
03
import org.apache.lucene.document.Document;
04
import org.apache.lucene.search.HitCollector;
05
import org.apache.lucene.search.IndexSearcher;
06
07
import java.io.IOException;
08
import java.util.Collections;
09
import java.util.HashMap;
10
import java.util.Map;
11
//
自定义
BookLinkCollector
的实现
,
比较简单
12
public class BookLinkCollector extends HitCollector {
13
private IndexSearcher searcher;
14
//
保存
URL
和
Title
的
Map
15
private HashMap documents = new HashMap();
16
17
public BookLinkCollector(IndexSearcher searcher) {
18
this.searcher = searcher;
19
}
20
21
//
实现的接口的方法
22
public void collect(int id, float score) {
23
try {
24
Document doc = searcher.doc(id);
25
documents.put(doc.get("url"), doc.get("title"));
26
System.out.println(doc.get("title") + ":" + score);
27
} catch (IOException e) {
28
// ignore
29
}
30
}
31
32
public Map getLinks() {
33
return Collections.unmodifiableMap(documents);
34
}
35
}
测试代码:
01
package lia.extsearch.hitcollector;
02
03
import lia.common.LiaTestCase;
04
import lia.extsearch.hitcollector.BookLinkCollector;
05
import org.apache.lucene.index.Term;
06
import org.apache.lucene.search.IndexSearcher;
07
import org.apache.lucene.search.TermQuery;
08
import org.apache.lucene.search.Hits;
09
10
import java.util.Map;
11
12
public class HitCollectorTest extends LiaTestCase {
13
14
public void testCollecting() throws Exception {
15
TermQuery query = new TermQuery(new Term("contents", "junit"));
16
IndexSearcher searcher = new IndexSearcher(directory);
17
18
// BookLinkCollector
需要一个参数
searcher
19
BookLinkCollector collector = new BookLinkCollector(searcher);
20
searcher.search(query, collector); //
搜索
21
22
Map linkMap = collector.getLinks();
23
//
测试
24
assertEquals("Java Development with Ant", linkMap
25
.get("http://www.manning.com/antbook"));
26
27
28
Hits hits = searcher.search(query);
29
dumpHits(hits);
30
31
searcher.close();
32
}
33
}
该实现是比较简单的,要进一步了解其用法请参考Lucene in Action 或者 我的Blog.
III.
自定义Filter的实现
有了上面实现的Sort代码 自定义实现Filter也是很简单的 只要实现Filter接口的一个方法就可以了 该方法如下:
public abstract BitSetbits(IndexReader reader)
throws IOException
Returns a BitSet with true for documents which should be permitted in search results, and false for those that should not.
来看个例子:
01
package lia.extsearch.filters;
02
03
import org.apache.lucene.index.IndexReader;
04
import org.apache.lucene.index.Term;
05
import org.apache.lucene.index.TermDocs;
06
import org.apache.lucene.search.Filter;
07
08
import java.io.IOException;
09
import java.util.BitSet;
10
11
import lia.extsearch.filters.SpecialsAccessor;
12
13
public class SpecialsFilter extends Filter {
14
//
访问
isbns
的接口
解耦
便于重用
15
private SpecialsAccessor accessor;
16
17
public SpecialsFilter(SpecialsAccessor accessor) {
18
this.accessor = accessor;
19
}
20
21
//
覆盖该方法
实现自定义
Filter
22
/**
23
* Returns a BitSet with true for documents which should be permitted in
24
* search results, and false for those that should not
25
*/
26
public BitSet bits(IndexReader reader) throws IOException {
27
BitSet bits = new BitSet(reader.maxDoc());
28
29
String[] isbns = accessor.isbns();
30
31
int[] docs = new int[1];
32
int[] freqs = new int[1];
33
34
for (int i = 0; i < isbns.length; i++) {
35
String isbn = isbns[i];
36
if (isbn != null) {
37
TermDocs termDocs = reader.termDocs(new Term("isbn", isbn));
38
int count = termDocs.read(docs, freqs);
39
if (count == 1) {
40
bits.set(docs[0]);
41
42
}
43
}
44
}
45
46
return bits;
47
}
48
49
public String toString() {
50
return "SpecialsFilter";
51
}
52
}
用到了如下接口
1
package lia.extsearch.filters;
2
3
//
定义一个取得过虑参考信息的接口
4
public interface SpecialsAccessor {
5
String[] isbns();
6
}
和
Mock Object
实现
01
package lia.extsearch.filters;
02
03
//
一个
Mock object
的实现
04
public class MockSpecialsAccessor implements SpecialsAccessor {
05
private String[] isbns;
06
07
public MockSpecialsAccessor(String[] isbns) {
08
this.isbns = isbns;
09
}
10
11
public String[] isbns() {
12
return isbns;
13
}
14
}
测试代码如下:
01
package lia.extsearch.filters;
02
03
import lia.common.LiaTestCase;
04
import org.apache.lucene.search.Filter;
05
import org.apache.lucene.search.Hits;
06
import org.apache.lucene.search.WildcardQuery;
07
import org.apache.lucene.search.FilteredQuery;
08
import org.apache.lucene.search.TermQuery;
09
import org.apache.lucene.search.BooleanQuery;
10
import org.apache.lucene.search.RangeQuery;
11
import org.apache.lucene.search.IndexSearcher;
12
import org.apache.lucene.search.Query;
13
import org.apache.lucene.index.Term;
14
15
//
测试自定义
Filter
16
public class SpecialsFilterTest extends LiaTestCase {
17
private Query allBooks;
18
19
private IndexSearcher searcher;
20
21
//
建立测试环境
22
protected void setUp() throws Exception {
23
super.setUp();
24
25
allBooks = new RangeQuery(new Term("pubmonth", "190001"), new Term(
26
"pubmonth", "200512"), true);
27
searcher = new IndexSearcher(directory);
28
}
29
30
//
测试
31
public void testCustomFilter() throws Exception {
32
String[] isbns = new String[] { "0060812451", "0465026567" };
33
34
SpecialsAccessor accessor = new MockSpecialsAccessor(isbns);
35
Filter filter = new SpecialsFilter(accessor);
36
Hits hits = searcher.search(allBooks, filter);
37
assertEquals("the specials", isbns.length, hits.length());
38
}
39
40
// Using the new FilteredQuery, though, you can apply a
41
// Filter to a particular query clause of a BooleanQuery.
42
// FilteredQuery
为
1.4
新加入的
详细情况请参考
Lucene in action
和
FilteredQuery
的
doc
43
public void testFilteredQuery() throws Exception {
44
String[] isbns = new String[] { "0854402624" }; // Steiner
45
46
SpecialsAccessor accessor = new MockSpecialsAccessor(isbns);
47
Filter filter = new SpecialsFilter(accessor);
48
49
WildcardQuery educationBooks = new WildcardQuery(new Term("category",
50
"*education*"));
51
FilteredQuery edBooksOnSpecial = new FilteredQuery(educationBooks,
52
filter);
53
54
TermQuery logoBooks = new TermQuery(new Term("subject", "logo"));
55
56
BooleanQuery logoOrEdBooks = new BooleanQuery();
57
logoOrEdBooks.add(logoBooks, false, false);
58
logoOrEdBooks.add(edBooksOnSpecial, false, false);
59
60
Hits hits = searcher.search(logoOrEdBooks);
61
System.out.println(logoOrEdBooks.toString());
62
assertEquals("Papert and Steiner", 2, hits.length());
63
}
64
}
明天看看 扩展
QueryParser
和
Lucene
的性能
看看
Lucene
到底有多块
!.
posted on 2007-01-05 10:27
Lansing 阅读(711)
评论(0) 编辑 收藏 所属分类:
搜索引擎