Apache
/
lucene
mirror of https://github.com/apache/lucene.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

LUCENE-1461: added FieldCacheRangeFilter, which speeds up creation of RangeFilter by using the FieldCache 

git-svn-id: https://svn.apache.org/repos/asf/lucene/java/trunk@721663 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Michael McCandless 2008-11-29 11:50:28 +00:00
parent 755e7aec14
commit f562594180
5 changed files with 595 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
CHANGES.txt
View File

@ -96,6 +96,15 @@ New features
Deprecated ConstantScoreRangeQuery (Mark Miller via Mike
McCandless)
7. LUCENE-1461: Added FieldCacheRangeFilter, a RangeFilter for
single-term fields that uses FieldCache to compute the filter. If
your field has a single term per document, and you need to create
many RangeFilters with varying lower/upper bounds, then this is
likely a much faster way to create the filters than RangeFilter.
However, it comes at the expense of added RAM consumption and
slower first-time usage due to populating the FieldCache. (Tim
Sturge via Mike McCandless)
Optimizations
1. LUCENE-1427: Fixed QueryWrapperFilter to not waste time computing

22
src/java/org/apache/lucene/search/FieldCache.java
View File

@ -38,7 +38,29 @@ public interface FieldCache {
/** Expert: Stores term text values and document ordering data. */
public static class StringIndex {
public int binarySearchLookup(String key) {
// this special case is the reason that Arrays.binarySearch() isn't useful.
if (key == null)
return 0;
int low = 1;
int high = lookup.length-1;
while (low <= high) {
int mid = (low + high) >> 1;
int cmp = lookup[mid].compareTo(key);
if (cmp < 0)
low = mid + 1;
else if (cmp > 0)
high = mid - 1;
else
return mid; // key found
}
return -(low + 1); // key not found.
}
/** All the term values, in natural order. */
public final String[] lookup;

182
src/java/org/apache/lucene/search/FieldCacheRangeFilter.java Normal file
View File

@ -0,0 +1,182 @@
package org.apache.lucene.search;
/**
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
import java.io.IOException;
import org.apache.lucene.index.IndexReader;
/**
* A range filter built on top of a cached single term field (in FieldCache).
*
* FieldCacheRangeFilter builds a single cache for the field the first time it is used.
*
* Each subsequent FieldCacheRangeFilter on the same field then reuses this cache,
* even if the range itself changes.
*
* This means that FieldCacheRangeFilter is much faster (sometimes more than 100x as fast)
* as building a RangeFilter (or ConstantScoreRangeQuery on a RangeFilter) for each query.
* However, if the range never changes it is slower (around 2x as slow) than building a
* CachingWrapperFilter on top of a single RangeFilter.
*
* As with all FieldCache based functionality, FieldCacheRangeFilter is only valid for
* fields which contain zero or one terms for each document. Thus it works on dates,
* prices and other single value fields but will not work on regular text fields. It is
* preferable to use an UN_TOKENIZED field to ensure that there is only a single term.
*
* Also, collation is done at the time the FieldCache is built; to change
* collation you need to override the getFieldCache() method to change the underlying cache.
*/
public class FieldCacheRangeFilter extends Filter {
private String field;
private String lowerVal;
private String upperVal;
private boolean includeLower;
private boolean includeUpper;
public FieldCacheRangeFilter(
String field,
String lowerVal,
String upperVal,
boolean includeLower,
boolean includeUpper) {
this.field = field;
this.lowerVal = lowerVal;
this.upperVal = upperVal;
this.includeLower = includeLower;
this.includeUpper = includeUpper;
}
public FieldCache getFieldCache() {
return FieldCache.DEFAULT;
}
public DocIdSet getDocIdSet(IndexReader reader) throws IOException {
return new RangeMultiFilterDocIdSet(getFieldCache().getStringIndex(reader, field));
}
public String toString() {
StringBuffer buffer = new StringBuffer();
buffer.append(field);
buffer.append(":");
buffer.append(includeLower ? "[" : "{");
if (null != lowerVal) {
buffer.append(lowerVal);
}
buffer.append("-");
if (null != upperVal) {
buffer.append(upperVal);
}
buffer.append(includeUpper ? "]" : "}");
return buffer.toString();
}
public boolean equals(Object o) {
if (this == o) return true;
if (!(o instanceof FieldCacheRangeFilter)) return false;
FieldCacheRangeFilter other = (FieldCacheRangeFilter) o;
if (!this.field.equals(other.field)
|| this.includeLower != other.includeLower
|| this.includeUpper != other.includeUpper
) { return false; }
if (this.lowerVal != null ? !this.lowerVal.equals(other.lowerVal) : other.lowerVal != null) return false;
if (this.upperVal != null ? !this.upperVal.equals(other.upperVal) : other.upperVal != null) return false;
return true;
}
public int hashCode() {
int h = field.hashCode();
h ^= lowerVal != null ? lowerVal.hashCode() : 550356204;
h = (h << 1) | (h >>> 31); // rotate to distinguish lower from upper
h ^= (upperVal != null ? (upperVal.hashCode()) : -1674416163);
h ^= (includeLower ? 1549299360 : -365038026)
^ (includeUpper ? 1721088258 : 1948649653);
return h;
}
protected class RangeMultiFilterDocIdSet extends DocIdSet {
private int inclusiveLowerPoint;
private int inclusiveUpperPoint;
private FieldCache.StringIndex fcsi;
public RangeMultiFilterDocIdSet(FieldCache.StringIndex fcsi) {
this.fcsi = fcsi;
initialize();
}
private void initialize() {
int lowerPoint = fcsi.binarySearchLookup(lowerVal);
if (includeLower && lowerPoint >= 0) {
inclusiveLowerPoint = lowerPoint;
} else if (lowerPoint >= 0) {
inclusiveLowerPoint = lowerPoint+1;
} else {
inclusiveLowerPoint = -lowerPoint-1;
}
int upperPoint = fcsi.binarySearchLookup(upperVal);
if (includeUpper && upperPoint >= 0) {
inclusiveUpperPoint = upperPoint;
} else if (upperPoint >= 0) {
inclusiveUpperPoint = upperPoint - 1;
} else {
inclusiveUpperPoint = -upperPoint - 2;
}
}
public DocIdSetIterator iterator() {
return new RangeMultiFilterIterator();
}
protected class RangeMultiFilterIterator extends DocIdSetIterator {
private int doc = -1;
public int doc() {
return doc;
}
public boolean next() {
try {
do {
doc++;
} while (fcsi.order[doc] > inclusiveUpperPoint
|| fcsi.order[doc] < inclusiveLowerPoint);
return true;
} catch (ArrayIndexOutOfBoundsException e) {
doc = Integer.MAX_VALUE;
return false;
}
}
public boolean skipTo(int target) {
try {
doc = target;
while (fcsi.order[doc] > inclusiveUpperPoint
|| fcsi.order[doc] < inclusiveLowerPoint) {
doc++;
}
return true;
} catch (ArrayIndexOutOfBoundsException e) {
doc = Integer.MAX_VALUE;
return false;
}
}
}
}
}

3
src/java/org/apache/lucene/search/RangeFilter.java
View File

@ -31,6 +31,9 @@ import java.text.Collator;
* This code borrows heavily from {@link RangeQuery}, but is implemented as a Filter
*
* </p>
*
* If you construct a large number of range filters with different ranges but on the
* same field, {@link FieldCacheRangeFilter} may have significantly better performance.
*/
public class RangeFilter extends Filter {

379
src/test/org/apache/lucene/search/TestFieldCacheRangeFilter.java Normal file
View File

@ -0,0 +1,379 @@
package org.apache.lucene.search;
/**
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
import java.io.IOException;
import java.text.Collator;
import java.util.Locale;
import org.apache.lucene.index.IndexReader;
import org.apache.lucene.index.Term;
import org.apache.lucene.index.IndexWriter;
import org.apache.lucene.analysis.SimpleAnalyzer;
import org.apache.lucene.document.Document;
import org.apache.lucene.document.Field;
import org.apache.lucene.store.RAMDirectory;
/**
* A basic 'positive' Unit test class for the RangeFilter class.
*
* <p>
* NOTE: at the moment, this class only tests for 'positive' results,
* it does not verify the results to ensure there are no 'false positives',
* nor does it adequately test 'negative' results. It also does not test
* that garbage in results in an Exception.
*/
public class TestFieldCacheRangeFilter extends BaseTestRangeFilter {
public TestFieldCacheRangeFilter(String name) {
super(name);
}
public TestFieldCacheRangeFilter() {
super();
}
public void testRangeFilterId() throws IOException {
IndexReader reader = IndexReader.open(signedIndex.index);
IndexSearcher search = new IndexSearcher(reader);
int medId = ((maxId - minId) / 2);
String minIP = pad(minId);
String maxIP = pad(maxId);
String medIP = pad(medId);
int numDocs = reader.numDocs();
assertEquals("num of docs", numDocs, 1+ maxId - minId);
ScoreDoc[] result;
Query q = new TermQuery(new Term("body","body"));
// test id, bounded on both ends
result = search.search(q,new RangeFilter("id",minIP,maxIP,T,T), numDocs).scoreDocs;
assertEquals("find all", numDocs, result.length);
result = search.search(q,new RangeFilter("id",minIP,maxIP,T,F), numDocs).scoreDocs;
assertEquals("all but last", numDocs-1, result.length);
result = search.search(q,new RangeFilter("id",minIP,maxIP,F,T), numDocs).scoreDocs;
assertEquals("all but first", numDocs-1, result.length);
result = search.search(q,new RangeFilter("id",minIP,maxIP,F,F), numDocs).scoreDocs;
assertEquals("all but ends", numDocs-2, result.length);
result = search.search(q,new RangeFilter("id",medIP,maxIP,T,T), numDocs).scoreDocs;
assertEquals("med and up", 1+ maxId-medId, result.length);
result = search.search(q,new RangeFilter("id",minIP,medIP,T,T), numDocs).scoreDocs;
assertEquals("up to med", 1+ medId-minId, result.length);
// unbounded id
result = search.search(q,new RangeFilter("id",minIP,null,T,F), numDocs).scoreDocs;
assertEquals("min and up", numDocs, result.length);
result = search.search(q,new RangeFilter("id",null,maxIP,F,T), numDocs).scoreDocs;
assertEquals("max and down", numDocs, result.length);
result = search.search(q,new RangeFilter("id",minIP,null,F,F), numDocs).scoreDocs;
assertEquals("not min, but up", numDocs-1, result.length);
result = search.search(q,new RangeFilter("id",null,maxIP,F,F), numDocs).scoreDocs;
assertEquals("not max, but down", numDocs-1, result.length);
result = search.search(q,new RangeFilter("id",medIP,maxIP,T,F), numDocs).scoreDocs;
assertEquals("med and up, not max", maxId-medId, result.length);
result = search.search(q,new RangeFilter("id",minIP,medIP,F,T), numDocs).scoreDocs;
assertEquals("not min, up to med", medId-minId, result.length);
// very small sets
result = search.search(q,new RangeFilter("id",minIP,minIP,F,F), numDocs).scoreDocs;
assertEquals("min,min,F,F", 0, result.length);
result = search.search(q,new RangeFilter("id",medIP,medIP,F,F), numDocs).scoreDocs;
assertEquals("med,med,F,F", 0, result.length);
result = search.search(q,new RangeFilter("id",maxIP,maxIP,F,F), numDocs).scoreDocs;
assertEquals("max,max,F,F", 0, result.length);
result = search.search(q,new RangeFilter("id",minIP,minIP,T,T), numDocs).scoreDocs;
assertEquals("min,min,T,T", 1, result.length);
result = search.search(q,new RangeFilter("id",null,minIP,F,T), numDocs).scoreDocs;
assertEquals("nul,min,F,T", 1, result.length);
result = search.search(q,new RangeFilter("id",maxIP,maxIP,T,T), numDocs).scoreDocs;
assertEquals("max,max,T,T", 1, result.length);
result = search.search(q,new RangeFilter("id",maxIP,null,T,F), numDocs).scoreDocs;
assertEquals("max,nul,T,T", 1, result.length);
result = search.search(q,new RangeFilter("id",medIP,medIP,T,T), numDocs).scoreDocs;
assertEquals("med,med,T,T", 1, result.length);
}
public void testRangeFilterIdCollating() throws IOException {
IndexReader reader = IndexReader.open(signedIndex.index);
IndexSearcher search = new IndexSearcher(reader);
Collator c = Collator.getInstance(Locale.ENGLISH);
int medId = ((maxId - minId) / 2);
String minIP = pad(minId);
String maxIP = pad(maxId);
String medIP = pad(medId);
int numDocs = reader.numDocs();
assertEquals("num of docs", numDocs, 1+ maxId - minId);
Hits result;
Query q = new TermQuery(new Term("body","body"));
// test id, bounded on both ends
result = search.search(q,new RangeFilter("id",minIP,maxIP,T,T,c));
assertEquals("find all", numDocs, result.length());
result = search.search(q,new RangeFilter("id",minIP,maxIP,T,F,c));
assertEquals("all but last", numDocs-1, result.length());
result = search.search(q,new RangeFilter("id",minIP,maxIP,F,T,c));
assertEquals("all but first", numDocs-1, result.length());
result = search.search(q,new RangeFilter("id",minIP,maxIP,F,F,c));
assertEquals("all but ends", numDocs-2, result.length());
result = search.search(q,new RangeFilter("id",medIP,maxIP,T,T,c));
assertEquals("med and up", 1+ maxId-medId, result.length());
result = search.search(q,new RangeFilter("id",minIP,medIP,T,T,c));
assertEquals("up to med", 1+ medId-minId, result.length());
// unbounded id
result = search.search(q,new RangeFilter("id",minIP,null,T,F,c));
assertEquals("min and up", numDocs, result.length());
result = search.search(q,new RangeFilter("id",null,maxIP,F,T,c));
assertEquals("max and down", numDocs, result.length());
result = search.search(q,new RangeFilter("id",minIP,null,F,F,c));
assertEquals("not min, but up", numDocs-1, result.length());
result = search.search(q,new RangeFilter("id",null,maxIP,F,F,c));
assertEquals("not max, but down", numDocs-1, result.length());
result = search.search(q,new RangeFilter("id",medIP,maxIP,T,F,c));
assertEquals("med and up, not max", maxId-medId, result.length());
result = search.search(q,new RangeFilter("id",minIP,medIP,F,T,c));
assertEquals("not min, up to med", medId-minId, result.length());
// very small sets
result = search.search(q,new RangeFilter("id",minIP,minIP,F,F,c));
assertEquals("min,min,F,F", 0, result.length());
result = search.search(q,new RangeFilter("id",medIP,medIP,F,F,c));
assertEquals("med,med,F,F", 0, result.length());
result = search.search(q,new RangeFilter("id",maxIP,maxIP,F,F,c));
assertEquals("max,max,F,F", 0, result.length());
result = search.search(q,new RangeFilter("id",minIP,minIP,T,T,c));
assertEquals("min,min,T,T", 1, result.length());
result = search.search(q,new RangeFilter("id",null,minIP,F,T,c));
assertEquals("nul,min,F,T", 1, result.length());
result = search.search(q,new RangeFilter("id",maxIP,maxIP,T,T,c));
assertEquals("max,max,T,T", 1, result.length());
result = search.search(q,new RangeFilter("id",maxIP,null,T,F,c));
assertEquals("max,nul,T,T", 1, result.length());
result = search.search(q,new RangeFilter("id",medIP,medIP,T,T,c));
assertEquals("med,med,T,T", 1, result.length());
}
public void testRangeFilterRand() throws IOException {
IndexReader reader = IndexReader.open(signedIndex.index);
IndexSearcher search = new IndexSearcher(reader);
String minRP = pad(signedIndex.minR);
String maxRP = pad(signedIndex.maxR);
int numDocs = reader.numDocs();
assertEquals("num of docs", numDocs, 1+ maxId - minId);
ScoreDoc[] result;
Query q = new TermQuery(new Term("body","body"));
// test extremes, bounded on both ends
result = search.search(q,new RangeFilter("rand",minRP,maxRP,T,T), numDocs).scoreDocs;
assertEquals("find all", numDocs, result.length);
result = search.search(q,new RangeFilter("rand",minRP,maxRP,T,F), numDocs).scoreDocs;
assertEquals("all but biggest", numDocs-1, result.length);
result = search.search(q,new RangeFilter("rand",minRP,maxRP,F,T), numDocs).scoreDocs;
assertEquals("all but smallest", numDocs-1, result.length);
result = search.search(q,new RangeFilter("rand",minRP,maxRP,F,F), numDocs).scoreDocs;
assertEquals("all but extremes", numDocs-2, result.length);
// unbounded
result = search.search(q,new RangeFilter("rand",minRP,null,T,F), numDocs).scoreDocs;
assertEquals("smallest and up", numDocs, result.length);
result = search.search(q,new RangeFilter("rand",null,maxRP,F,T), numDocs).scoreDocs;
assertEquals("biggest and down", numDocs, result.length);
result = search.search(q,new RangeFilter("rand",minRP,null,F,F), numDocs).scoreDocs;
assertEquals("not smallest, but up", numDocs-1, result.length);
result = search.search(q,new RangeFilter("rand",null,maxRP,F,F), numDocs).scoreDocs;
assertEquals("not biggest, but down", numDocs-1, result.length);
// very small sets
result = search.search(q,new RangeFilter("rand",minRP,minRP,F,F), numDocs).scoreDocs;
assertEquals("min,min,F,F", 0, result.length);
result = search.search(q,new RangeFilter("rand",maxRP,maxRP,F,F), numDocs).scoreDocs;
assertEquals("max,max,F,F", 0, result.length);
result = search.search(q,new RangeFilter("rand",minRP,minRP,T,T), numDocs).scoreDocs;
assertEquals("min,min,T,T", 1, result.length);
result = search.search(q,new RangeFilter("rand",null,minRP,F,T), numDocs).scoreDocs;
assertEquals("nul,min,F,T", 1, result.length);
result = search.search(q,new RangeFilter("rand",maxRP,maxRP,T,T), numDocs).scoreDocs;
assertEquals("max,max,T,T", 1, result.length);
result = search.search(q,new RangeFilter("rand",maxRP,null,T,F), numDocs).scoreDocs;
assertEquals("max,nul,T,T", 1, result.length);
}
public void testRangeFilterRandCollating() throws IOException {
// using the unsigned index because collation seems to ignore hyphens
IndexReader reader = IndexReader.open(unsignedIndex.index);
IndexSearcher search = new IndexSearcher(reader);
Collator c = Collator.getInstance(Locale.ENGLISH);
String minRP = pad(unsignedIndex.minR);
String maxRP = pad(unsignedIndex.maxR);
int numDocs = reader.numDocs();
assertEquals("num of docs", numDocs, 1+ maxId - minId);
Hits result;
Query q = new TermQuery(new Term("body","body"));
// test extremes, bounded on both ends
result = search.search(q,new RangeFilter("rand",minRP,maxRP,T,T,c));
assertEquals("find all", numDocs, result.length());
result = search.search(q,new RangeFilter("rand",minRP,maxRP,T,F,c));
assertEquals("all but biggest", numDocs-1, result.length());
result = search.search(q,new RangeFilter("rand",minRP,maxRP,F,T,c));
assertEquals("all but smallest", numDocs-1, result.length());
result = search.search(q,new RangeFilter("rand",minRP,maxRP,F,F,c));
assertEquals("all but extremes", numDocs-2, result.length());
// unbounded
result = search.search(q,new RangeFilter("rand",minRP,null,T,F,c));
assertEquals("smallest and up", numDocs, result.length());
result = search.search(q,new RangeFilter("rand",null,maxRP,F,T,c));
assertEquals("biggest and down", numDocs, result.length());
result = search.search(q,new RangeFilter("rand",minRP,null,F,F,c));
assertEquals("not smallest, but up", numDocs-1, result.length());
result = search.search(q,new RangeFilter("rand",null,maxRP,F,F,c));
assertEquals("not biggest, but down", numDocs-1, result.length());
// very small sets
result = search.search(q,new RangeFilter("rand",minRP,minRP,F,F,c));
assertEquals("min,min,F,F", 0, result.length());
result = search.search(q,new RangeFilter("rand",maxRP,maxRP,F,F,c));
assertEquals("max,max,F,F", 0, result.length());
result = search.search(q,new RangeFilter("rand",minRP,minRP,T,T,c));
assertEquals("min,min,T,T", 1, result.length());
result = search.search(q,new RangeFilter("rand",null,minRP,F,T,c));
assertEquals("nul,min,F,T", 1, result.length());
result = search.search(q,new RangeFilter("rand",maxRP,maxRP,T,T,c));
assertEquals("max,max,T,T", 1, result.length());
result = search.search(q,new RangeFilter("rand",maxRP,null,T,F,c));
assertEquals("max,nul,T,T", 1, result.length());
}
public void testFarsi() throws Exception {
/* build an index */
RAMDirectory farsiIndex = new RAMDirectory();
IndexWriter writer = new IndexWriter(farsiIndex, new SimpleAnalyzer(), T,
IndexWriter.MaxFieldLength.LIMITED);
Document doc = new Document();
doc.add(new Field("content","\u0633\u0627\u0628",
Field.Store.YES, Field.Index.UN_TOKENIZED));
doc.add(new Field("body", "body",
Field.Store.YES, Field.Index.UN_TOKENIZED));
writer.addDocument(doc);
writer.optimize();
writer.close();
IndexReader reader = IndexReader.open(farsiIndex);
IndexSearcher search = new IndexSearcher(reader);
Query q = new TermQuery(new Term("body","body"));
// Neither Java 1.4.2 nor 1.5.0 has Farsi Locale collation available in
// RuleBasedCollator. However, the Arabic Locale seems to order the Farsi
// characters properly.
Collator collator = Collator.getInstance(new Locale("ar"));
// Unicode order would include U+0633 in [ U+062F - U+0698 ], but Farsi
// orders the U+0698 character before the U+0633 character, so the single
// index Term below should NOT be returned by a RangeFilter with a Farsi
// Collator (or an Arabic one for the case when Farsi is not supported).
Hits result = search.search
(q, new RangeFilter("content", "\u062F", "\u0698", T, T, collator));
assertEquals("The index Term should not be included.", 0, result.length());
result = search.search
(q, new RangeFilter("content", "\u0633", "\u0638", T, T, collator));
assertEquals("The index Term should be included.", 1, result.length());
search.close();
}
}