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

LUCENE-7214: Remove two-phase iteration from LatLonPoint.newDistanceQuery

This commit is contained in:
Robert Muir 2016-04-13 15:56:32 -04:00
parent 689e9664c5
commit fc19c99e0e
2 changed files with 18 additions and 64 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
lucene/CHANGES.txt
View File

@ -39,9 +39,6 @@ Optimizations
* LUCENE-7071: Reduce bytes copying in OfflineSorter, giving ~10%
speedup on merging 2D LatLonPoint values (Mike McCandless)
* LUCENE-7099: LatLonPoint's newDistanceQuery supports two-phase
iteration. (Robert Muir)
* LUCENE-7105: Optimize LatLonPoint's newDistanceQuery. (Robert Muir)
* LUCENE-7109: LatLonPoint's newPolygonQuery supports two-phase

79
lucene/sandbox/src/java/org/apache/lucene/document/LatLonPointDistanceQuery.java
View File

@ -18,13 +18,12 @@ package org.apache.lucene.document;
import java.io.IOException;
import org.apache.lucene.geo.GeoUtils;
import org.apache.lucene.geo.Rectangle;
import org.apache.lucene.index.DocValues;
import org.apache.lucene.index.FieldInfo;
import org.apache.lucene.index.LeafReader;
import org.apache.lucene.index.LeafReaderContext;
import org.apache.lucene.index.PointValues;
import org.apache.lucene.index.SortedNumericDocValues;
import org.apache.lucene.index.PointValues.IntersectVisitor;
import org.apache.lucene.index.PointValues.Relation;
import org.apache.lucene.search.ConstantScoreScorer;
@ -34,15 +33,10 @@ import org.apache.lucene.search.DocIdSetIterator;
import org.apache.lucene.search.IndexSearcher;
import org.apache.lucene.search.Query;
import org.apache.lucene.search.Scorer;
import org.apache.lucene.search.TwoPhaseIterator;
import org.apache.lucene.search.Weight;
import org.apache.lucene.geo.GeoUtils;
import org.apache.lucene.util.BitSet;
import org.apache.lucene.util.DocIdSetBuilder;
import org.apache.lucene.util.FixedBitSet;
import org.apache.lucene.util.NumericUtils;
import org.apache.lucene.util.SloppyMath;
import org.apache.lucene.util.SparseFixedBitSet;
import org.apache.lucene.util.StringHelper;
import static org.apache.lucene.geo.GeoEncodingUtils.decodeLatitude;
@ -104,7 +98,7 @@ final class LatLonPointDistanceQuery extends Query {
}
// compute a maximum partial haversin: unless our box is crazy, we can use this bound
// to reject edge cases faster in matches()
// to reject edge cases faster in visit()
final double maxPartialDistance;
if (box.maxLon - longitude < 90 && longitude - box.minLon < 90) {
maxPartialDistance = Math.max(SloppyMath.haversinSortKey(latitude, longitude, latitude, box.maxLon),
@ -132,16 +126,9 @@ final class LatLonPointDistanceQuery extends Query {
}
LatLonPoint.checkCompatible(fieldInfo);
// approximation (postfiltering has not yet been applied)
// matching docids
DocIdSetBuilder result = new DocIdSetBuilder(reader.maxDoc());
// subset of documents that need no postfiltering, this is purely an optimization
final BitSet preApproved;
// dumb heuristic: if the field is really sparse, use a sparse impl
if (values.getDocCount(field) * 100L < reader.maxDoc()) {
preApproved = new SparseFixedBitSet(reader.maxDoc());
} else {
preApproved = new FixedBitSet(reader.maxDoc());
}
values.intersect(field,
new IntersectVisitor() {
@Override
@ -152,14 +139,11 @@ final class LatLonPointDistanceQuery extends Query {
@Override
public void visit(int docID) {
result.add(docID);
preApproved.set(docID);
}
@Override
public void visit(int docID, byte[] packedValue) {
// we bounds check individual values, as subtrees may cross, but we are being sent the values anyway:
// this reduces the amount of docvalues fetches (improves approximation)
// bounding box check
if (StringHelper.compare(Integer.BYTES, packedValue, 0, maxLat, 0) > 0 ||
StringHelper.compare(Integer.BYTES, packedValue, 0, minLat, 0) < 0) {
// latitude out of bounding box range
@ -173,13 +157,23 @@ final class LatLonPointDistanceQuery extends Query {
return;
}
result.add(docID);
double docLatitude = decodeLatitude(packedValue, 0);
double docLongitude = decodeLongitude(packedValue, Integer.BYTES);
// first check the partial distance, if its more than that, it can't be <= radiusMeters
double h1 = SloppyMath.haversinSortKey(latitude, longitude, docLatitude, docLongitude);
if (h1 <= maxPartialDistance) {
// fully confirm with part 2:
if (SloppyMath.haversinMeters(h1) <= radiusMeters) {
result.add(docID);
}
}
}
// algorithm: we create a bounding box (two bounding boxes if we cross the dateline).
// 1. check our bounding box(es) first. if the subtree is entirely outside of those, bail.
// 2. check if the subtree is disjoint. it may cross the bounding box but not intersect with circle
// 3. see if the subtree is fully contained. if the subtree is enormous along the x axis, wrapping half way around the world, etc: then this can't work, just go to step 3.
// 3. see if the subtree is fully contained. if the subtree is enormous along the x axis, wrapping half way around the world, etc: then this can't work, just go to step 4.
// 4. recurse naively (subtrees crossing over circle edge)
@Override
public Relation compare(byte[] minPackedValue, byte[] maxPackedValue) {
@ -231,44 +225,7 @@ final class LatLonPointDistanceQuery extends Query {
if (disi == null) {
return null;
}
// return two-phase iterator using docvalues to postfilter candidates
SortedNumericDocValues docValues = DocValues.getSortedNumeric(reader, field);
TwoPhaseIterator iterator = new TwoPhaseIterator(disi) {
@Override
public boolean matches() throws IOException {
int docId = disi.docID();
if (preApproved.get(docId)) {
return true;
} else {
docValues.setDocument(docId);
int count = docValues.count();
for (int i = 0; i < count; i++) {
long encoded = docValues.valueAt(i);
double docLatitude = decodeLatitude((int)(encoded >> 32));
double docLongitude = decodeLongitude((int)(encoded & 0xFFFFFFFF));
// first check the partial distance, if its more than that, it can't be <= radiusMeters
double h1 = SloppyMath.haversinSortKey(latitude, longitude, docLatitude, docLongitude);
if (h1 > maxPartialDistance) {
continue;
}
// fully confirm with part 2:
if (SloppyMath.haversinMeters(h1) <= radiusMeters) {
return true;
}
}
return false;
}
}
@Override
public float matchCost() {
return 20; // TODO: make this fancier
}
};
return new ConstantScoreScorer(this, score(), iterator);
return new ConstantScoreScorer(this, score(), disi);
}
};
}