使用Lucene-Spatial实现集成地理位置的全文检索

转自：/article/1456209.html

Lucene通过Spatial包提供了对基于地理位置的全文检索的支持，最典型的应用场景就是：“搜索中关村附近1公里内的火锅店，并按远近排序”。使用Lucene-Spatial添加对地理位置的支持，和之前普通文本搜索主要有两点区别：

1. 将坐标信息转化为笛卡尔层，建立索引

[java]
view plaincopyprint?

private void indexLocation(Document document, JSONObject jo)
throws Exception {

double longitude = jo.getDouble("longitude");
double latitude = jo.getDouble("latitude");

document.add(new Field("lat", NumericUtils
.doubleToPrefixCoded(latitude), Field.Store.YES,
Field.Index.NOT_ANALYZED));
document.add(new Field("lng", NumericUtils
.doubleToPrefixCoded(longitude), Field.Store.YES,
Field.Index.NOT_ANALYZED));

for (int tier = startTier; tier <= endTier; tier++) {
ctp = new CartesianTierPlotter(tier, projector,
CartesianTierPlotter.DEFALT_FIELD_PREFIX);
final double boxId = ctp.getTierBoxId(latitude, longitude);
document.add(new Field(ctp.getTierFieldName(), NumericUtils
.doubleToPrefixCoded(boxId), Field.Store.YES,
Field.Index.NOT_ANALYZED_NO_NORMS));
}
}

2. 搜索时，指定使用DistanceQueryFilter

[java]
view plaincopyprint?

DistanceQueryBuilder dq = new DistanceQueryBuilder(latitude,
longitude, miles, "lat", "lng",
CartesianTierPlotter.DEFALT_FIELD_PREFIX, true, startTier,
endTier);
DistanceFieldComparatorSource dsort = new DistanceFieldComparatorSource(
dq.getDistanceFilter());
Sort sort = new Sort(new SortField("geo_distance", dsort));

下面是基于Lucene3.2.0和JUnit4.8.2的完整代码。

[html]
view plaincopyprint?

<dependencies>
<dependency>
<groupId>junit</groupId>
<artifactId>junit</artifactId>
<version>4.8.2</version>
<type>jar</type>
<scope>test</scope>
</dependency>
<dependency>
<groupId>org.apache.lucene</groupId>
<artifactId>lucene-core</artifactId>
<version>3.2.0</version>
<type>jar</type>
<scope>compile</scope>
</dependency>
<dependency>
<groupId>org.apache.lucene</groupId>
<artifactId>lucene-spatial</artifactId>
<version>3.2.0</version>
<type>jar</type>
<scope>compile</scope>
</dependency>
<dependency>
<groupId>org.json</groupId>
<artifactId>json</artifactId>
<version>20100903</version>
<type>jar</type>
<scope>compile</scope>
</dependency>
</dependencies>

首先准备测试用的数据：

[plain]
view plaincopyprint?

{"id":12,"title":"时尚码头美容美发热烫特价","longitude":116.3838183,"latitude":39.9629015}
{"id":17,"title":"审美个人美容美发套餐","longitude":116.386564,"latitude":39.966102}
{"id":23,"title":"海底捞吃300送300","longitude":116.38629,"latitude":39.9629573}
{"id":26,"title":"仅98元！享原价335元李老爹","longitude":116.3846175,"latitude":39.9629125}
{"id":29,"title":"都美造型烫染美发护理套餐","longitude":116.38629,"latitude":39.9629573}
{"id":30,"title":"仅售55元！原价80元的老舍茶馆相声下午场","longitude":116.0799914,"latitude":39.9655391}
{"id":33,"title":"仅售55元！原价80元的新笑声客栈早场","longitude":116.0799914,"latitude":39.9655391}
{"id":34,"title":"仅售39元（红色礼盒）！原价80元的平谷桃","longitude":116.0799914,"latitude":39.9655391}
{"id":46,"title":"仅售38元！原价180元地质礼堂白雪公主","longitude":116.0799914,"latitude":39.9655391}
{"id":49,"title":"仅99元！享原价342.7元自助餐","longitude":116.0799914,"latitude":39.9655391}
{"id":58,"title":"桑海教育暑期学生报名培训九折优惠券","longitude":116.0799914,"latitude":39.9655391}
{"id":59,"title":"全国发货：仅29元！贝玲妃超模粉红高光光","longitude":116.0799914,"latitude":39.9655391}
{"id":65,"title":"海之屿生态水族用品店抵用券","longitude":116.0799914,"latitude":39.9655391}
{"id":67,"title":"小区东门时尚烫染个人护理美发套餐","longitude":116.3799914,"latitude":39.9655391}
{"id":74,"title":"《郭德纲相声专辑》CD套装","longitude":116.0799914,"latitude":39.9655391}

根据上面的测试数据，编写测试用例，分别搜索坐标（116.3838183, 39.9629015）3千米以内的“美发”和全部内容，分别得到的结果应该是4条和6条。

[java]
view plaincopyprint?

import static org.junit.Assert.assertEquals;
import static org.junit.Assert.fail;

import java.util.List;

import org.junit.Test;

public class LuceneSpatialTest {

private static LuceneSpatial spatialSearcher = new LuceneSpatial();

@Test
public void testSearch() {
try {
long start = System.currentTimeMillis();
List<String> results = spatialSearcher.search("美发", 116.3838183, 39.9629015, 3.0);
System.out.println(results.size()
+ "个匹配结果，共耗时 "
+ (System.currentTimeMillis() - start) + "毫秒。\n");
assertEquals(4, results.size());
} catch (Exception e) {
fail("Exception occurs...");
e.printStackTrace();
}
}

@Test
public void testSearchWithoutKeyword() {
try {
long start = System.currentTimeMillis();
List<String> results = spatialSearcher.search(null, 116.3838183, 39.9629015, 3.0);
System.out.println( results.size()
+ "个匹配结果，共耗时 "
+ (System.currentTimeMillis() - start) + "毫秒.\n");
assertEquals(6, results.size());
} catch (Exception e) {
fail("Exception occurs...");
e.printStackTrace();
}
}
}

下面是LuceneSpatial类，在构造函数中初始化变量和创建索引：

[java]
view plaincopyprint?

public class LuceneSpatial {

private Analyzer analyzer;
private IndexWriter writer;
private FSDirectory indexDirectory;
private IndexSearcher indexSearcher;
private IndexReader indexReader;
private String indexPath = "c:/lucene-spatial";

// Spatial
private IProjector projector;
private CartesianTierPlotter ctp;
public static final double RATE_MILE_TO_KM = 1.609344; //英里和公里的比率
public static final String LAT_FIELD = "lat";
public static final String LON_FIELD = "lng";
private static final double MAX_RANGE = 15.0; // 索引支持的最大范围，单位是千米
private static final double MIN_RANGE = 3.0; // 索引支持的最小范围，单位是千米
private int startTier;
private int endTier;

public LuceneSpatial() {
try {
init();
} catch (Exception e) {
e.printStackTrace();
}
}

private void init() throws Exception {
initializeSpatialOptions();

analyzer = new StandardAnalyzer(Version.LUCENE_32);

File path = new File(indexPath);

boolean isNeedCreateIndex = false;

if (path.exists() && !path.isDirectory())
throw new Exception("Specified path is not a directory");

if (!path.exists()) {
path.mkdirs();
isNeedCreateIndex = true;
}

indexDirectory = FSDirectory.open(new File(indexPath));

//建立索引
if (isNeedCreateIndex) {
IndexWriterConfig indexWriterConfig = new IndexWriterConfig(
Version.LUCENE_32, analyzer);
indexWriterConfig.setOpenMode(OpenMode.CREATE_OR_APPEND);
writer = new IndexWriter(indexDirectory, indexWriterConfig);
buildIndex();
}

indexReader = IndexReader.open(indexDirectory, true);
indexSearcher = new IndexSearcher(indexReader);

}

@SuppressWarnings("deprecation")
private void initializeSpatialOptions() {
projector = new SinusoidalProjector();
ctp = new CartesianTierPlotter(0, projector,
CartesianTierPlotter.DEFALT_FIELD_PREFIX);
startTier = ctp.bestFit(MAX_RANGE / RATE_MILE_TO_KM);
endTier = ctp.bestFit(MIN_RANGE / RATE_MILE_TO_KM);
}

private int mile2Meter(double miles) {
double dMeter = miles * RATE_MILE_TO_KM * 1000;

return (int) dMeter;
}

private double km2Mile(double km) {
return km / RATE_MILE_TO_KM;
}

创建索引的具体实现：

[java]
view plaincopyprint?

private void buildIndex() {
BufferedReader br = null;
try {
//逐行添加测试数据到索引中，测试数据文件和源文件在同一个目录下
br = new BufferedReader(new InputStreamReader(
LuceneSpatial.class.getResourceAsStream("data")));
String line = null;
while ((line = br.readLine()) != null) {
index(new JSONObject(line));
}

writer.commit();
} catch (Exception e) {
e.printStackTrace();
} finally {
if (br != null) {
try {
br.close();
} catch (IOException e) {
e.printStackTrace();
}
}
}
}

private void index(JSONObject jo) throws Exception {
Document doc = new Document();

doc.add(new Field("id", jo.getString("id"), Field.Store.YES,
Field.Index.ANALYZED));

doc.add(new Field("title", jo.getString("title"), Field.Store.YES,
Field.Index.ANALYZED));

//将位置信息添加到索引中
indexLocation(doc, jo);

writer.addDocument(doc);
}

private void indexLocation(Document document, JSONObject jo)
throws Exception {

double longitude = jo.getDouble("longitude");
double latitude = jo.getDouble("latitude");

document.add(new Field("lat", NumericUtils
.doubleToPrefixCoded(latitude), Field.Store.YES,
Field.Index.NOT_ANALYZED));
document.add(new Field("lng", NumericUtils
.doubleToPrefixCoded(longitude), Field.Store.YES,
Field.Index.NOT_ANALYZED));

for (int tier = startTier; tier <= endTier; tier++) {
ctp = new CartesianTierPlotter(tier, projector,
CartesianTierPlotter.DEFALT_FIELD_PREFIX);
final double boxId = ctp.getTierBoxId(latitude, longitude);
document.add(new Field(ctp.getTierFieldName(), NumericUtils
.doubleToPrefixCoded(boxId), Field.Store.YES,
Field.Index.NOT_ANALYZED_NO_NORMS));
}
}

搜索的具体实现：

[java]
view plaincopyprint?

public List<String> search(String keyword, double longitude,
double latitude, double range) throws Exception {
List<String> result = new ArrayList<String>();

double miles = km2Mile(range);

DistanceQueryBuilder dq = new DistanceQueryBuilder(latitude,
longitude, miles, "lat", "lng",
CartesianTierPlotter.DEFALT_FIELD_PREFIX, true, startTier,
endTier);

//按照距离排序
DistanceFieldComparatorSource dsort = new DistanceFieldComparatorSource(
dq.getDistanceFilter());
Sort sort = new Sort(new SortField("geo_distance", dsort));

Query query = buildQuery(keyword);

//搜索结果
TopDocs hits = indexSearcher.search(query, dq.getFilter(),
Integer.MAX_VALUE, sort);
//获得各条结果相对应的距离
Map<Integer, Double> distances = dq.getDistanceFilter()
.getDistances();

for (int i = 0; i < hits.totalHits; i++) {
final int docID = hits.scoreDocs[i].doc;

final Document doc = indexSearcher.doc(docID);

final StringBuilder builder = new StringBuilder();
builder.append("找到了: ")
.append(doc.get("title"))
.append("， 距离: ")
.append(mile2Meter(distances.get(docID)))
.append("米。");
System.out.println(builder.toString());

result.add(builder.toString());
}

return result;
}

private Query buildQuery(String keyword) throws Exception {
//如果没有指定关键字，则返回范围内的所有结果
if (keyword == null || keyword.isEmpty()) {
return new MatchAllDocsQuery();
}
QueryParser parser = new QueryParser(Version.LUCENE_32, "title",
analyzer);

parser.setDefaultOperator(Operator.AND);

return parser.parse(keyword.toString());
}

执行测试用例，可以得到下面的结果：

[plain]
view plaincopyprint?

找到了: 时尚码头美容美发热烫特价， 距离: 0米。
找到了: 都美造型烫染美发护理套餐， 距离: 210米。
找到了: 审美个人美容美发套餐， 距离: 426米。
找到了: 小区东门时尚烫染个人护理美发套餐， 距离: 439米。
4个匹配结果，共耗时 119毫秒。

找到了: 时尚码头美容美发热烫特价， 距离: 0米。
找到了: 仅98元！享原价335元李老爹， 距离: 68米。
找到了: 海底捞吃300送300， 距离: 210米。
找到了: 都美造型烫染美发护理套餐， 距离: 210米。
找到了: 审美个人美容美发套餐， 距离: 426米。
找到了: 小区东门时尚烫染个人护理美发套餐， 距离: 439米。
6个匹配结果，共耗时 3毫秒.

参考文献：
Lucene-Spatial的原理介绍：http://www.nsshutdown.com/projects/lucene/whitepaper/locallucene.htm

GeoHash：http://en.wikipedia.org/wiki/Geohash

两篇示例（其中大部分代码就来自于这里）：

Spatial
search with Lucene