1
|
/**
|
2
|
* Copyright (C) 2011 EDIT
|
3
|
* European Distributed Institute of Taxonomy
|
4
|
* http://www.e-taxonomy.eu
|
5
|
*
|
6
|
* The contents of this file are subject to the Mozilla Public License Version 1.1
|
7
|
* See LICENSE.TXT at the top of this package for the full license terms.
|
8
|
*/
|
9
|
package eu.etaxonomy.cdm.api.service.search;
|
10
|
|
11
|
import java.io.IOException;
|
12
|
import java.util.Arrays;
|
13
|
import java.util.Collection;
|
14
|
|
15
|
import org.apache.log4j.Logger;
|
16
|
import org.apache.lucene.analysis.Analyzer;
|
17
|
import org.apache.lucene.queryparser.classic.ParseException;
|
18
|
import org.apache.lucene.search.BooleanClause.Occur;
|
19
|
import org.apache.lucene.search.BooleanQuery;
|
20
|
import org.apache.lucene.search.BooleanQuery.Builder;
|
21
|
import org.apache.lucene.search.IndexSearcher;
|
22
|
import org.apache.lucene.search.MultiCollector;
|
23
|
import org.apache.lucene.search.Query;
|
24
|
import org.apache.lucene.search.Sort;
|
25
|
import org.apache.lucene.search.SortField;
|
26
|
import org.apache.lucene.search.TopDocs;
|
27
|
import org.apache.lucene.search.grouping.GroupDocs;
|
28
|
import org.apache.lucene.search.grouping.SearchGroup;
|
29
|
import org.apache.lucene.search.grouping.TopGroups;
|
30
|
import org.apache.lucene.search.grouping.term.TermAllGroupsCollector;
|
31
|
import org.apache.lucene.search.grouping.term.TermFirstPassGroupingCollector;
|
32
|
import org.apache.lucene.search.grouping.term.TermSecondPassGroupingCollector;
|
33
|
import org.apache.lucene.util.BytesRef;
|
34
|
|
35
|
import eu.etaxonomy.cdm.model.common.CdmBase;
|
36
|
import eu.etaxonomy.cdm.model.description.DescriptionElementBase;
|
37
|
import eu.etaxonomy.cdm.model.description.TextData;
|
38
|
import eu.etaxonomy.cdm.model.taxon.Taxon;
|
39
|
import eu.etaxonomy.cdm.model.taxon.TaxonBase;
|
40
|
|
41
|
/**
|
42
|
*
|
43
|
* @author Andreas Kohlbecker
|
44
|
\* @since Dec 21, 2011
|
45
|
*
|
46
|
*/
|
47
|
public class LuceneSearch {
|
48
|
|
49
|
protected String groupByField = "id";
|
50
|
|
51
|
public final static String ID_FIELD = "id";
|
52
|
|
53
|
public static final Logger logger = Logger.getLogger(LuceneSearch.class);
|
54
|
|
55
|
protected ILuceneIndexToolProvider toolProvider;
|
56
|
|
57
|
protected IndexSearcher searcher;
|
58
|
|
59
|
protected SortField[] sortFields;
|
60
|
|
61
|
private Class<? extends CdmBase> directorySelectClass;
|
62
|
|
63
|
private BooleanQuery filter = null;
|
64
|
|
65
|
protected Class<? extends CdmBase> getDirectorySelectClass() {
|
66
|
return pushAbstractBaseTypeDown(directorySelectClass);
|
67
|
}
|
68
|
|
69
|
/**
|
70
|
* classFilter
|
71
|
*/
|
72
|
protected Class<? extends CdmBase> cdmTypeRestriction;
|
73
|
|
74
|
|
75
|
public Class<? extends CdmBase> getCdmTypRestriction() {
|
76
|
return cdmTypeRestriction;
|
77
|
}
|
78
|
|
79
|
/**
|
80
|
* @return the filter
|
81
|
*/
|
82
|
public BooleanQuery getFilter() {
|
83
|
return filter;
|
84
|
}
|
85
|
|
86
|
/**
|
87
|
* @param filter the filter to set
|
88
|
*/
|
89
|
public void setFilter(BooleanQuery filter) {
|
90
|
this.filter = filter;
|
91
|
}
|
92
|
|
93
|
/**
|
94
|
* Sets the Class to use as filter criterion, in case the supplied Class equals the
|
95
|
* <code>directorySelectClass</code> the Class is set to <code>null</code>
|
96
|
* @param clazz
|
97
|
*/
|
98
|
public void setCdmTypRestriction(Class<? extends CdmBase> clazz) {
|
99
|
|
100
|
/*
|
101
|
* NOTE:
|
102
|
* we must not use the getter of directorySelectClass
|
103
|
* since we need the abstract base classes here!!!!
|
104
|
*/
|
105
|
if(clazz != null && clazz.equals(directorySelectClass)){
|
106
|
clazz = null;
|
107
|
}
|
108
|
this.cdmTypeRestriction = clazz;
|
109
|
}
|
110
|
|
111
|
/**
|
112
|
* The MAX_HITS_ALLOWED value must be one less than Integer.MAX_VALUE
|
113
|
* otherwise PriorityQueue will produce an exception since it
|
114
|
* will always add 1 to the maxhits so Integer.MAX_VALUE
|
115
|
* would become Integer.MIN_VALUE
|
116
|
*/
|
117
|
public final int MAX_HITS_ALLOWED = 10000;
|
118
|
|
119
|
protected BooleanQuery query;
|
120
|
|
121
|
protected String[] highlightFields = new String[0];
|
122
|
|
123
|
private int maxDocsPerGroup = 10;
|
124
|
|
125
|
|
126
|
public int getMaxDocsPerGroup() {
|
127
|
return maxDocsPerGroup;
|
128
|
}
|
129
|
|
130
|
public void setMaxDocsPerGroup(int maxDocsPerGroup) {
|
131
|
this.maxDocsPerGroup = maxDocsPerGroup;
|
132
|
}
|
133
|
|
134
|
/**
|
135
|
* @param session
|
136
|
*/
|
137
|
public LuceneSearch(ILuceneIndexToolProvider toolProvider, Class<? extends CdmBase> directorySelectClass) {
|
138
|
this.toolProvider = toolProvider;
|
139
|
this.directorySelectClass = directorySelectClass;
|
140
|
}
|
141
|
|
142
|
/**
|
143
|
* @param session
|
144
|
*/
|
145
|
public LuceneSearch(ILuceneIndexToolProvider toolProvider, String groupByField, Class<? extends CdmBase> directorySelectClass) {
|
146
|
this.toolProvider = toolProvider;
|
147
|
this.directorySelectClass = directorySelectClass;
|
148
|
this.groupByField = groupByField;
|
149
|
}
|
150
|
|
151
|
/**
|
152
|
* TODO the abstract base class DescriptionElementBase can not be used, so
|
153
|
* we are using an arbitrary subclass to find the DirectoryProvider, future
|
154
|
* versions of hibernate search my allow using abstract base classes see
|
155
|
* {@link http://stackoverflow.com/questions/492184/how-do-you-find-all-subclasses-of-a-given-class-in-java}
|
156
|
*
|
157
|
* @param type must not be null
|
158
|
* @return
|
159
|
*/
|
160
|
private Class<? extends CdmBase> pushAbstractBaseTypeDown(Class<? extends CdmBase> type) {
|
161
|
Class<? extends CdmBase> returnType = type;
|
162
|
if (type.equals(DescriptionElementBase.class)) {
|
163
|
returnType = TextData.class;
|
164
|
}
|
165
|
if (type.equals(TaxonBase.class)) {
|
166
|
returnType = Taxon.class;
|
167
|
}
|
168
|
// if (type.equals(TaxonName.class)) {
|
169
|
// returnType = NonViralName.class;
|
170
|
// }
|
171
|
return returnType;
|
172
|
}
|
173
|
|
174
|
protected LuceneSearch() {
|
175
|
|
176
|
}
|
177
|
|
178
|
/**
|
179
|
* @return
|
180
|
*/
|
181
|
public IndexSearcher getSearcher() {
|
182
|
if(searcher == null){
|
183
|
searcher = new IndexSearcher(toolProvider.getIndexReaderFor(directorySelectClass));
|
184
|
// searcher.setDefaultFieldSortScoring(true, true);
|
185
|
}
|
186
|
return searcher;
|
187
|
}
|
188
|
|
189
|
/**
|
190
|
* Convenience method which delegated the call to the available
|
191
|
* {@link ILuceneIndexToolProvider#getAnalyzerFor(Class)} method.
|
192
|
*
|
193
|
* @return the Analyzer suitable for the <code>directorySelectClass</code>
|
194
|
* of the LuceneSearch
|
195
|
*/
|
196
|
public Analyzer getAnalyzer() {
|
197
|
return toolProvider.getAnalyzerFor(directorySelectClass);
|
198
|
}
|
199
|
|
200
|
/**
|
201
|
* @param luceneQueryString
|
202
|
* @param cdmTypeRestriction the type as additional filter criterion
|
203
|
* @param pageSize if the page size is null or in an invalid range it will be set to MAX_HITS_ALLOWED
|
204
|
* @param pageNumber a 0-based index of the page to return, will default to 0 if null or negative.
|
205
|
* @return
|
206
|
* @throws ParseException
|
207
|
* @throws IOException
|
208
|
*/
|
209
|
public TopGroups<BytesRef> executeSearch(String luceneQueryString, Integer pageSize, Integer pageNumber) throws ParseException, IOException {
|
210
|
|
211
|
Query luceneQuery = parse(luceneQueryString);
|
212
|
setQuery(luceneQuery);
|
213
|
return executeSearch(pageSize, pageNumber);
|
214
|
}
|
215
|
|
216
|
/**
|
217
|
* @param luceneQueryString
|
218
|
* @return
|
219
|
* @throws ParseException
|
220
|
*/
|
221
|
public Query parse(String luceneQueryString) throws ParseException {
|
222
|
logger.debug("luceneQueryString to be parsed: " + luceneQueryString);
|
223
|
Query luceneQuery = toolProvider.getQueryParserFor(directorySelectClass, false).parse(luceneQueryString);
|
224
|
return luceneQuery;
|
225
|
}
|
226
|
|
227
|
/**
|
228
|
* @param maxNoOfHits
|
229
|
* @return
|
230
|
* @throws IOException
|
231
|
*/
|
232
|
public TopDocs executeSearch(int maxNoOfHits) throws IOException {
|
233
|
BooleanQuery fullQuery = expandQuery();
|
234
|
logger.info("lucene query string to be parsed: " + fullQuery.toString());
|
235
|
return getSearcher().search(fullQuery, maxNoOfHits, Sort.RELEVANCE, true, true);
|
236
|
|
237
|
}
|
238
|
/**
|
239
|
* @param pageSize if the page size is null or in an invalid range it will be set to MAX_HITS_ALLOWED
|
240
|
* @param pageNumber a 0-based index of the page to return, will default to 0 if null or negative.
|
241
|
* @return
|
242
|
* @throws ParseException
|
243
|
* @throws IOException
|
244
|
*/
|
245
|
public TopGroups<BytesRef> executeSearch(Integer pageSize, Integer pageNumber) throws ParseException, IOException {
|
246
|
|
247
|
|
248
|
if(pageNumber == null || pageNumber < 0){
|
249
|
pageNumber = 0;
|
250
|
}
|
251
|
if(pageSize == null || pageSize <= 0 || pageSize > MAX_HITS_ALLOWED){
|
252
|
pageSize = MAX_HITS_ALLOWED;
|
253
|
logger.info("limiting pageSize to MAX_HITS_ALLOWED = " + MAX_HITS_ALLOWED + " items");
|
254
|
}
|
255
|
|
256
|
BooleanQuery fullQuery = expandQuery();
|
257
|
logger.info("final query: " + fullQuery.toString());
|
258
|
|
259
|
int offset = pageNumber * pageSize;
|
260
|
int limit = (pageNumber + 1) * pageSize - 1 ;
|
261
|
logger.debug("start: " + offset + "; limit:" + limit);
|
262
|
|
263
|
// sorting
|
264
|
Sort groupSort = null;
|
265
|
Sort withinGroupSort = Sort.RELEVANCE;
|
266
|
if(sortFields != null && sortFields.length > 0){
|
267
|
groupSort = new Sort(sortFields);
|
268
|
} else {
|
269
|
groupSort = Sort.RELEVANCE; // == SortField.FIELD_SCORE !!
|
270
|
}
|
271
|
|
272
|
// perform the search (needs two passes for grouping)
|
273
|
if(logger.isDebugEnabled()){
|
274
|
logger.debug("Grouping: sortFields=" + Arrays.toString(sortFields) + ", groupByField=" + groupByField +
|
275
|
", groupSort=" + groupSort + ", withinGroupSort=" + withinGroupSort + ", limit=" + limit + ", maxDocsPerGroup="+ maxDocsPerGroup);
|
276
|
}
|
277
|
// - first pass
|
278
|
TermFirstPassGroupingCollector firstPassCollector = new TermFirstPassGroupingCollector(groupByField, groupSort, limit);
|
279
|
|
280
|
getSearcher().search(fullQuery, firstPassCollector);
|
281
|
Collection<SearchGroup<BytesRef>> topGroups = firstPassCollector.getTopGroups(0, true); // no offset here since we need the first item for the max score
|
282
|
|
283
|
if (topGroups == null) {
|
284
|
return null;
|
285
|
}
|
286
|
// - flags for second pass
|
287
|
boolean getScores = false;
|
288
|
boolean getMaxScores = true;
|
289
|
if(groupSort.getSort()[0] != SortField.FIELD_SCORE){
|
290
|
getMaxScores = false;
|
291
|
// see inner class TopGroupsWithMaxScore
|
292
|
logger.error("Fist sort field must be SortField.FIELD_SCORE otherwise the max score value will not be correct! MaxScore calculation will be skipped");
|
293
|
}
|
294
|
boolean fillFields = true;
|
295
|
TermAllGroupsCollector allGroupsCollector = new TermAllGroupsCollector(groupByField);
|
296
|
TermSecondPassGroupingCollector secondPassCollector = new TermSecondPassGroupingCollector(
|
297
|
groupByField, topGroups, groupSort, withinGroupSort, maxDocsPerGroup , getScores,
|
298
|
getMaxScores, fillFields
|
299
|
);
|
300
|
getSearcher().search(fullQuery, MultiCollector.wrap(secondPassCollector, allGroupsCollector));
|
301
|
|
302
|
TopGroups<BytesRef> groupsResult = secondPassCollector.getTopGroups(0); // no offset here since we need the first item for the max score
|
303
|
|
304
|
// --- set the max score for the group results
|
305
|
|
306
|
// get max score from very first result
|
307
|
float maxScore = groupsResult.groups[0].maxScore;
|
308
|
|
309
|
if(logger.isDebugEnabled()){
|
310
|
logger.debug("TopGroups: maxScore=" + maxScore + ", offset=" + offset +
|
311
|
", totalGroupCount=" + allGroupsCollector.getGroupCount() +
|
312
|
", totalGroupedHitCount=" + groupsResult.totalGroupedHitCount);
|
313
|
}
|
314
|
|
315
|
TopGroups<BytesRef> newTopGroups;
|
316
|
if(offset > 0){
|
317
|
GroupDocs<BytesRef>[] newGroupDocs = new GroupDocs[groupsResult.groups.length - offset];
|
318
|
for(int i = offset; i < groupsResult.groups.length; i++){
|
319
|
newGroupDocs[i - offset] = groupsResult.groups[i];
|
320
|
}
|
321
|
newTopGroups = new TopGroups<BytesRef>(
|
322
|
groupsResult.groupSort,
|
323
|
groupsResult.withinGroupSort,
|
324
|
groupsResult.totalHitCount,
|
325
|
groupsResult.totalGroupedHitCount,
|
326
|
newGroupDocs,
|
327
|
maxScore);
|
328
|
} else {
|
329
|
newTopGroups = groupsResult;
|
330
|
}
|
331
|
TopGroups<BytesRef> topGroupsWithMaxScore = new TopGroups<BytesRef>(newTopGroups, allGroupsCollector.getGroupCount());
|
332
|
// --- done with max score for the group results
|
333
|
|
334
|
return topGroupsWithMaxScore;
|
335
|
}
|
336
|
|
337
|
/**
|
338
|
* expands the query by adding a type restriction if the
|
339
|
* <code>cdmTypeRestriction</code> is not <code>NULL</code>
|
340
|
* and adds the <code>filter</code> as Boolean query
|
341
|
* clause with {@link Occur#FILTER}
|
342
|
*/
|
343
|
protected BooleanQuery expandQuery() {
|
344
|
BooleanQuery fullQuery = null;
|
345
|
Builder fullQueryBuilder = null;
|
346
|
|
347
|
if(cdmTypeRestriction != null){
|
348
|
fullQueryBuilder = QueryFactory.addTypeRestriction(query, cdmTypeRestriction);
|
349
|
}
|
350
|
|
351
|
if(filter != null) {
|
352
|
if(fullQueryBuilder == null) {
|
353
|
fullQueryBuilder = new Builder();
|
354
|
fullQueryBuilder.add(this.query, Occur.MUST);
|
355
|
}
|
356
|
fullQueryBuilder.add(filter, Occur.FILTER);
|
357
|
}
|
358
|
|
359
|
if(fullQueryBuilder != null) {
|
360
|
fullQuery = fullQueryBuilder.build();
|
361
|
} else {
|
362
|
fullQuery = this.query;
|
363
|
}
|
364
|
|
365
|
logger.debug("expandedQuery: " + fullQuery.toString());
|
366
|
return fullQuery;
|
367
|
}
|
368
|
|
369
|
public void setQuery(Query query) {
|
370
|
if( query instanceof BooleanQuery) {
|
371
|
this.query = (BooleanQuery)query;
|
372
|
} else {
|
373
|
Builder builder = new Builder();
|
374
|
this.query = builder.add(query, Occur.MUST).build();
|
375
|
}
|
376
|
}
|
377
|
|
378
|
public BooleanQuery getQuery() {
|
379
|
return query;
|
380
|
}
|
381
|
|
382
|
public BooleanQuery getExpandedQuery() {
|
383
|
expandQuery();
|
384
|
return query;
|
385
|
}
|
386
|
|
387
|
public SortField[] getSortFields() {
|
388
|
return sortFields;
|
389
|
}
|
390
|
|
391
|
public void setSortFields(SortField[] sortFields) {
|
392
|
this.sortFields = sortFields;
|
393
|
}
|
394
|
|
395
|
public void setHighlightFields(String[] textFieldNamesAsArray) {
|
396
|
this.highlightFields = textFieldNamesAsArray;
|
397
|
}
|
398
|
|
399
|
public String[] getHighlightFields() {
|
400
|
return this.highlightFields;
|
401
|
}
|
402
|
|
403
|
}
|