LUCENE-5565: Refactor SpatialPrefixTree/Cell to not use Strings. Other misc API changes to these classes too.

git-svn-id: https://svn.apache.org/repos/asf/lucene/dev/trunk@1584793 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
David Wayne Smiley 2014-04-04 16:27:41 +00:00
parent 85642f59dd
commit 6414781969
12 changed files with 396 additions and 233 deletions

View File

@ -190,6 +190,8 @@ API Changes
* LUCENE-5543: Remove/deprecate Directory.fileExists (Mike McCandless) * LUCENE-5543: Remove/deprecate Directory.fileExists (Mike McCandless)
* LUCENE-5565: Refactor SpatialPrefixTree/Cell to not use Strings. (David Smiley)
Optimizations Optimizations
* LUCENE-5468: HunspellStemFilter uses 10 to 100x less RAM. It also loads * LUCENE-5468: HunspellStemFilter uses 10 to 100x less RAM. It also loads

View File

@ -25,7 +25,6 @@ import org.apache.lucene.spatial.prefix.tree.Cell;
import org.apache.lucene.spatial.prefix.tree.SpatialPrefixTree; import org.apache.lucene.spatial.prefix.tree.SpatialPrefixTree;
import org.apache.lucene.util.Bits; import org.apache.lucene.util.Bits;
import org.apache.lucene.util.BytesRef; import org.apache.lucene.util.BytesRef;
import org.apache.lucene.util.StringHelper;
import java.io.IOException; import java.io.IOException;
import java.util.Iterator; import java.util.Iterator;
@ -117,7 +116,7 @@ public abstract class AbstractVisitingPrefixTreeFilter extends AbstractPrefixTre
protected final boolean hasIndexedLeaves;//if false then we can skip looking for them protected final boolean hasIndexedLeaves;//if false then we can skip looking for them
private VNode curVNode;//current pointer, derived from query shape private VNode curVNode;//current pointer, derived from query shape
private BytesRef curVNodeTerm = new BytesRef();//curVNode.cell's term. private BytesRef curVNodeTerm = new BytesRef();//curVNode.cell's term, without leaf
private Cell scanCell; private Cell scanCell;
private BytesRef thisTerm;//the result of termsEnum.term() private BytesRef thisTerm;//the result of termsEnum.term()
@ -171,8 +170,7 @@ public abstract class AbstractVisitingPrefixTreeFilter extends AbstractPrefixTre
} }
//Seek to curVNode's cell (or skip if termsEnum has moved beyond) //Seek to curVNode's cell (or skip if termsEnum has moved beyond)
curVNodeTerm.bytes = curVNode.cell.getTokenBytes(); curVNode.cell.getTokenBytesNoLeaf(curVNodeTerm);
curVNodeTerm.length = curVNodeTerm.bytes.length;
int compare = thisTerm.compareTo(curVNodeTerm); int compare = thisTerm.compareTo(curVNodeTerm);
if (compare > 0) { if (compare > 0) {
// leap frog (termsEnum is beyond where we would otherwise seek) // leap frog (termsEnum is beyond where we would otherwise seek)
@ -215,7 +213,7 @@ public abstract class AbstractVisitingPrefixTreeFilter extends AbstractPrefixTre
if (hasIndexedLeaves && cell.getLevel() != 0) { if (hasIndexedLeaves && cell.getLevel() != 0) {
//If the next indexed term just adds a leaf marker ('+') to cell, //If the next indexed term just adds a leaf marker ('+') to cell,
// then add all of those docs // then add all of those docs
assert StringHelper.startsWith(thisTerm, curVNodeTerm);//TODO refactor to use method on curVNode.cell assert curVNode.cell.isWithin(curVNodeTerm, thisTerm);
scanCell = grid.getCell(thisTerm.bytes, thisTerm.offset, thisTerm.length, scanCell); scanCell = grid.getCell(thisTerm.bytes, thisTerm.offset, thisTerm.length, scanCell);
if (scanCell.getLevel() == cell.getLevel() && scanCell.isLeaf()) { if (scanCell.getLevel() == cell.getLevel() && scanCell.isLeaf()) {
visitLeaf(scanCell); visitLeaf(scanCell);
@ -265,7 +263,7 @@ public abstract class AbstractVisitingPrefixTreeFilter extends AbstractPrefixTre
*/ */
protected void scan(int scanDetailLevel) throws IOException { protected void scan(int scanDetailLevel) throws IOException {
for (; for (;
thisTerm != null && StringHelper.startsWith(thisTerm, curVNodeTerm);//TODO refactor to use method on curVNode.cell thisTerm != null && curVNode.cell.isWithin(curVNodeTerm, thisTerm);
thisTerm = termsEnum.next()) { thisTerm = termsEnum.next()) {
scanCell = grid.getCell(thisTerm.bytes, thisTerm.offset, thisTerm.length, scanCell); scanCell = grid.getCell(thisTerm.bytes, thisTerm.offset, thisTerm.length, scanCell);

View File

@ -0,0 +1,183 @@
package org.apache.lucene.spatial.prefix;
/*
* 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 org.apache.lucene.analysis.TokenStream;
import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
import org.apache.lucene.analysis.tokenattributes.TermToBytesRefAttribute;
import org.apache.lucene.spatial.prefix.tree.Cell;
import org.apache.lucene.util.Attribute;
import org.apache.lucene.util.AttributeImpl;
import org.apache.lucene.util.AttributeReflector;
import org.apache.lucene.util.AttributeSource;
import org.apache.lucene.util.BytesRef;
import java.io.IOException;
import java.util.Iterator;
/**
* A TokenStream used internally by {@link org.apache.lucene.spatial.prefix.PrefixTreeStrategy}.
*
* This is highly modelled after {@link org.apache.lucene.analysis.NumericTokenStream}.
*
* If there is demand for it to be public; it could be made to be.
*
* @lucene.internal
*/
class CellTokenStream extends TokenStream {
private interface CellTermAttribute extends Attribute {
Cell getCell();
void setCell(Cell cell);
//TODO one day deprecate this once we have better encodings
boolean getOmitLeafByte();
void setOmitLeafByte(boolean b);
}
// just a wrapper to prevent adding CTA
private static final class CellAttributeFactory extends AttributeSource.AttributeFactory {
private final AttributeSource.AttributeFactory delegate;
CellAttributeFactory(AttributeSource.AttributeFactory delegate) {
this.delegate = delegate;
}
@Override
public AttributeImpl createAttributeInstance(Class<? extends Attribute> attClass) {
if (CharTermAttribute.class.isAssignableFrom(attClass))
throw new IllegalArgumentException("CellTokenStream does not support CharTermAttribute.");
return delegate.createAttributeInstance(attClass);
}
}
private static final class CellTermAttributeImpl extends AttributeImpl
implements CellTermAttribute, TermToBytesRefAttribute {
private BytesRef bytes = new BytesRef();
private Cell cell;
private boolean omitLeafByte;//false by default (whether there's a leaf byte or not)
@Override
public Cell getCell() {
return cell;
}
@Override
public boolean getOmitLeafByte() {
return omitLeafByte;
}
@Override
public void setCell(Cell cell) {
this.cell = cell;
omitLeafByte = false;//reset
}
@Override
public void setOmitLeafByte(boolean b) {
omitLeafByte = b;
}
@Override
public void clear() {
// this attribute has no contents to clear!
// we keep it untouched as it's fully controlled by outer class.
}
@Override
public void copyTo(AttributeImpl target) {
final CellTermAttribute a = (CellTermAttribute) target;
a.setCell(cell);
a.setOmitLeafByte(omitLeafByte);
}
@Override
public int fillBytesRef() {
if (omitLeafByte)
cell.getTokenBytesNoLeaf(bytes);
else
cell.getTokenBytes(bytes);
return bytes.hashCode();
}
@Override
public BytesRef getBytesRef() {
return bytes;
}
@Override
public void reflectWith(AttributeReflector reflector) {
fillBytesRef();
reflector.reflect(TermToBytesRefAttribute.class, "bytes", BytesRef.deepCopyOf(bytes));
}
}
public CellTokenStream() {
this(AttributeFactory.DEFAULT_ATTRIBUTE_FACTORY);
}
public CellTokenStream(AttributeFactory factory) {
super(new CellAttributeFactory(factory));
}
public CellTokenStream setCells(Iterator<Cell> iter) {
this.iter = iter;
return this;
}
@Override
public void reset() throws IOException {
if (iter == null)
throw new IllegalStateException("call setCells() before usage");
cellAtt.setCell(null);
cellAtt.setOmitLeafByte(false);
}
/** Outputs the token of a cell, and if its a leaf, outputs it again with the leaf byte. */
@Override
public final boolean incrementToken() {
if (iter == null)
throw new IllegalStateException("call setCells() before usage");
// this will only clear all other attributes in this TokenStream
clearAttributes();
if (cellAtt.getOmitLeafByte()) {
cellAtt.setOmitLeafByte(false);
return true;
}
//get next
if (!iter.hasNext())
return false;
cellAtt.setCell(iter.next());
if (cellAtt.getCell().isLeaf())
cellAtt.setOmitLeafByte(true);
return true;
}
{
addAttributeImpl(new CellTermAttributeImpl());//because non-public constructor
}
//members
private final CellTermAttribute cellAtt = addAttribute(CellTermAttribute.class);
//TODO support position increment, and TypeAttribute
private Iterator<Cell> iter = null; // null means not initialized
}

View File

@ -83,7 +83,7 @@ public class ContainsPrefixTreeFilter extends AbstractPrefixTreeFilter {
super(context, acceptDocs); super(context, acceptDocs);
} }
BytesRef termBytes = new BytesRef(); BytesRef termBytes = new BytesRef();//no leaf
Cell nextCell;//see getLeafDocs Cell nextCell;//see getLeafDocs
/** This is the primary algorithm; recursive. Returns null if finds none. */ /** This is the primary algorithm; recursive. Returns null if finds none. */
@ -130,16 +130,15 @@ public class ContainsPrefixTreeFilter extends AbstractPrefixTreeFilter {
} }
private boolean seekExact(Cell cell) throws IOException { private boolean seekExact(Cell cell) throws IOException {
assert new BytesRef(cell.getTokenBytes()).compareTo(termBytes) > 0; assert cell.getTokenBytesNoLeaf(null).compareTo(termBytes) > 0;
termBytes.bytes = cell.getTokenBytes(); cell.getTokenBytesNoLeaf(termBytes);
termBytes.length = termBytes.bytes.length;
if (termsEnum == null) if (termsEnum == null)
return false; return false;
return termsEnum.seekExact(termBytes); return termsEnum.seekExact(termBytes);
} }
private SmallDocSet getDocs(Cell cell, Bits acceptContains) throws IOException { private SmallDocSet getDocs(Cell cell, Bits acceptContains) throws IOException {
assert new BytesRef(cell.getTokenBytes()).equals(termBytes); assert cell.getTokenBytesNoLeaf(null).equals(termBytes);
return collectDocs(acceptContains); return collectDocs(acceptContains);
} }
@ -147,7 +146,7 @@ public class ContainsPrefixTreeFilter extends AbstractPrefixTreeFilter {
private Cell lastLeaf = null;//just for assertion private Cell lastLeaf = null;//just for assertion
private SmallDocSet getLeafDocs(Cell leafCell, Bits acceptContains) throws IOException { private SmallDocSet getLeafDocs(Cell leafCell, Bits acceptContains) throws IOException {
assert new BytesRef(leafCell.getTokenBytes()).equals(termBytes); assert leafCell.getTokenBytesNoLeaf(null).equals(termBytes);
assert ! leafCell.equals(lastLeaf);//don't call for same leaf again assert ! leafCell.equals(lastLeaf);//don't call for same leaf again
lastLeaf = leafCell; lastLeaf = leafCell;

View File

@ -46,7 +46,7 @@ public class PointPrefixTreeFieldCacheProvider extends ShapeFieldCacheProvider<P
protected Point readShape(BytesRef term) { protected Point readShape(BytesRef term) {
scanCell = grid.getCell(term.bytes, term.offset, term.length, scanCell); scanCell = grid.getCell(term.bytes, term.offset, term.length, scanCell);
if (scanCell.getLevel() == grid.getMaxLevels() && !scanCell.isLeaf()) if (scanCell.getLevel() == grid.getMaxLevels() && !scanCell.isLeaf())
return scanCell.getCenter(); return scanCell.getShape().getCenter();
return null; return null;
} }
} }

View File

@ -19,8 +19,6 @@ package org.apache.lucene.spatial.prefix;
import com.spatial4j.core.shape.Point; import com.spatial4j.core.shape.Point;
import com.spatial4j.core.shape.Shape; import com.spatial4j.core.shape.Shape;
import org.apache.lucene.analysis.TokenStream;
import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
import org.apache.lucene.document.Field; import org.apache.lucene.document.Field;
import org.apache.lucene.document.FieldType; import org.apache.lucene.document.FieldType;
import org.apache.lucene.index.FieldInfo; import org.apache.lucene.index.FieldInfo;
@ -31,7 +29,6 @@ import org.apache.lucene.spatial.prefix.tree.SpatialPrefixTree;
import org.apache.lucene.spatial.query.SpatialArgs; import org.apache.lucene.spatial.query.SpatialArgs;
import org.apache.lucene.spatial.util.ShapeFieldCacheDistanceValueSource; import org.apache.lucene.spatial.util.ShapeFieldCacheDistanceValueSource;
import java.util.Iterator;
import java.util.List; import java.util.List;
import java.util.Map; import java.util.Map;
import java.util.concurrent.ConcurrentHashMap; import java.util.concurrent.ConcurrentHashMap;
@ -125,13 +122,12 @@ public abstract class PrefixTreeStrategy extends SpatialStrategy {
public Field[] createIndexableFields(Shape shape, double distErr) { public Field[] createIndexableFields(Shape shape, double distErr) {
int detailLevel = grid.getLevelForDistance(distErr); int detailLevel = grid.getLevelForDistance(distErr);
// note: maybe CellTokenStream should do this line, but it doesn't matter and it would create extra
// coupling
List<Cell> cells = grid.getCells(shape, detailLevel, true, simplifyIndexedCells);//intermediates cells List<Cell> cells = grid.getCells(shape, detailLevel, true, simplifyIndexedCells);//intermediates cells
//TODO is CellTokenStream supposed to be re-used somehow? see Uwe's comments:
// http://code.google.com/p/lucene-spatial-playground/issues/detail?id=4
Field field = new Field(getFieldName(), Field field = new Field(getFieldName(),
new CellTokenStream(cells.iterator()), FIELD_TYPE); new CellTokenStream().setCells(cells.iterator()), FIELD_TYPE);
return new Field[]{field}; return new Field[]{field};
} }
@ -146,41 +142,6 @@ public abstract class PrefixTreeStrategy extends SpatialStrategy {
FIELD_TYPE.freeze(); FIELD_TYPE.freeze();
} }
/** Outputs the tokenString of a cell, and if its a leaf, outputs it again with the leaf byte. */
final static class CellTokenStream extends TokenStream {
private final CharTermAttribute termAtt = addAttribute(CharTermAttribute.class);
private Iterator<Cell> iter = null;
public CellTokenStream(Iterator<Cell> tokens) {
this.iter = tokens;
}
CharSequence nextTokenStringNeedingLeaf = null;
@Override
public boolean incrementToken() {
clearAttributes();
if (nextTokenStringNeedingLeaf != null) {
termAtt.append(nextTokenStringNeedingLeaf);
termAtt.append((char) Cell.LEAF_BYTE);
nextTokenStringNeedingLeaf = null;
return true;
}
if (iter.hasNext()) {
Cell cell = iter.next();
CharSequence token = cell.getTokenString();
termAtt.append(token);
if (cell.isLeaf())
nextTokenStringNeedingLeaf = token;
return true;
}
return false;
}
}
@Override @Override
public ValueSource makeDistanceValueSource(Point queryPoint, double multiplier) { public ValueSource makeDistanceValueSource(Point queryPoint, double multiplier) {
PointPrefixTreeFieldCacheProvider p = provider.get( getFieldName() ); PointPrefixTreeFieldCacheProvider p = provider.get( getFieldName() );

View File

@ -61,7 +61,7 @@ public class TermQueryPrefixTreeStrategy extends PrefixTreeStrategy {
BytesRef[] terms = new BytesRef[cells.size()]; BytesRef[] terms = new BytesRef[cells.size()];
int i = 0; int i = 0;
for (Cell cell : cells) { for (Cell cell : cells) {
terms[i++] = new BytesRef(cell.getTokenString());//TODO use cell.getTokenBytes() terms[i++] = cell.getTokenBytesNoLeaf(null);
} }
return new TermsFilter(getFieldName(), terms); return new TermsFilter(getFieldName(), terms);
} }

View File

@ -20,6 +20,8 @@ package org.apache.lucene.spatial.prefix.tree;
import com.spatial4j.core.shape.Point; import com.spatial4j.core.shape.Point;
import com.spatial4j.core.shape.Shape; import com.spatial4j.core.shape.Shape;
import com.spatial4j.core.shape.SpatialRelation; import com.spatial4j.core.shape.SpatialRelation;
import org.apache.lucene.util.BytesRef;
import org.apache.lucene.util.StringHelper;
import java.util.ArrayList; import java.util.ArrayList;
import java.util.Collection; import java.util.Collection;
@ -27,74 +29,45 @@ import java.util.Collections;
import java.util.List; import java.util.List;
/** /**
* Represents a grid cell. These are not necessarily thread-safe, although new * Represents a grid cell. These are not necessarily thread-safe, although calling {@link #getShape()} will
* Cell("") (world cell) must be. * sufficiently prepare it to be so, if needed.
* *
* @lucene.experimental * @lucene.experimental
*/ */
public abstract class Cell implements Comparable<Cell> { public abstract class Cell {
public static final byte LEAF_BYTE = '+';//NOTE: must sort before letters & numbers
/* private static final byte LEAF_BYTE = '+';//NOTE: must sort before letters & numbers
Holds a byte[] and/or String representation of the cell. Both are lazy constructed from the other.
Neither contains the trailing leaf byte. //Arguably we could simply use a BytesRef, using an extra Object.
*/
private byte[] bytes; private byte[] bytes;
private int b_off; private int b_off;
private int b_len; private int b_len;
private String token;//this is the only part of equality
/** /**
* When set via getSubCells(filter), it is the relationship between this cell * When set via getSubCells(filter), it is the relationship between this cell
* and the given shape filter. * and the given shape filter. Doesn't participate in shape equality.
*/ */
protected SpatialRelation shapeRel; protected SpatialRelation shapeRel;
/** /** Warning: Refers to the same bytes (no copy). If {@link #setLeaf()} is subsequently called then it
* Always false for points. Otherwise, indicate no further sub-cells are going * may modify bytes. */
* to be provided because shapeRel is WITHIN or maxLevels or a detailLevel is
* hit.
*/
protected boolean leaf;
protected Cell(String token) {
this.token = token;
if (token.length() > 0 && token.charAt(token.length() - 1) == (char) LEAF_BYTE) {
this.token = token.substring(0, token.length() - 1);
setLeaf();
}
if (getLevel() == 0)
getShape();//ensure any lazy instantiation completes to make this threadsafe
}
protected Cell(byte[] bytes, int off, int len) { protected Cell(byte[] bytes, int off, int len) {
this.bytes = bytes; this.bytes = bytes;
this.b_off = off; this.b_off = off;
this.b_len = len; this.b_len = len;
b_fixLeaf();
} }
/** Warning: Refers to the same bytes (no copy). If {@link #setLeaf()} is subsequently called then it
* may modify bytes. */
public void reset(byte[] bytes, int off, int len) { public void reset(byte[] bytes, int off, int len) {
assert getLevel() != 0; assert getLevel() != 0;
token = null;
shapeRel = null; shapeRel = null;
this.bytes = bytes; this.bytes = bytes;
this.b_off = off; this.b_off = off;
this.b_len = len; this.b_len = len;
b_fixLeaf();
} }
private void b_fixLeaf() { protected abstract SpatialPrefixTree getGrid();
//note that non-point shapes always have the maxLevels cell set with setLeaf
if (bytes[b_off + b_len - 1] == LEAF_BYTE) {
b_len--;
setLeaf();
} else {
leaf = false;
}
}
public SpatialRelation getShapeRel() { public SpatialRelation getShapeRel() {
return shapeRel; return shapeRel;
@ -105,47 +78,68 @@ public abstract class Cell implements Comparable<Cell> {
* further cells with this prefix for the shape (always true at maxLevels). * further cells with this prefix for the shape (always true at maxLevels).
*/ */
public boolean isLeaf() { public boolean isLeaf() {
return leaf; return (b_len > 0 && bytes[b_off + b_len - 1] == LEAF_BYTE);
} }
/** Note: not supported at level 0. */ /** Modifies the bytes to reflect that this is a leaf. Warning: never invoke from a cell
* initialized to reference the same bytes from termsEnum, which should be treated as immutable.
* Note: not supported at level 0. */
public void setLeaf() { public void setLeaf() {
assert getLevel() != 0; assert getLevel() != 0;
leaf = true; if (isLeaf())
} return;
//if isn't big enough, we have to copy
/** if (bytes.length < b_off + b_len) {
* Note: doesn't contain a trailing leaf byte. //hopefully this copying doesn't happen too much (DWS: I checked and it doesn't seem to happen)
*/ byte[] copy = new byte[b_len + 1];
public String getTokenString() { System.arraycopy(bytes, b_off, copy, 0, b_len);
if (token == null) { copy[b_len++] = LEAF_BYTE;
token = new String(bytes, b_off, b_len, SpatialPrefixTree.UTF8); bytes = copy;
}
return token;
}
/**
* Note: doesn't contain a trailing leaf byte.
*/
public byte[] getTokenBytes() {
if (bytes != null) {
if (b_off != 0 || b_len != bytes.length) {
throw new IllegalStateException("Not supported if byte[] needs to be recreated.");
}
} else {
bytes = token.getBytes(SpatialPrefixTree.UTF8);
b_off = 0; b_off = 0;
b_len = bytes.length; } else {
bytes[b_off + b_len++] = LEAF_BYTE;
} }
return bytes;
} }
/**
* Returns the bytes for this cell.
* The result param is used to save object allocation, though it's bytes aren't used.
* @param result where the result goes, or null to create new
*/
public BytesRef getTokenBytes(BytesRef result) {
if (result == null)
result = new BytesRef();
result.bytes = bytes;
result.offset = b_off;
result.length = b_len;
return result;
}
/**
* Returns the bytes for this cell, without leaf set. The bytes should sort before any
* cells that have the leaf set for the spatial location.
* The result param is used to save object allocation, though it's bytes aren't used.
* @param result where the result goes, or null to create new
*/
public BytesRef getTokenBytesNoLeaf(BytesRef result) {
result = getTokenBytes(result);
if (isLeaf())
result.length--;
return result;
}
/** Level 0 is the world (and has no parent), from then on a higher level means a smaller
* cell than the level before it.
*/
public int getLevel() { public int getLevel() {
return token != null ? token.length() : b_len; return isLeaf() ? b_len - 1 : b_len;
} }
//TODO add getParent() and update some algorithms to use this? /** Gets the parent cell that contains this one. Don't call on the world cell. */
//public Cell getParent(); public Cell getParent() {
assert getLevel() > 0;
return getGrid().getCell(bytes, b_off, b_len - (isLeaf() ? 2 : 1));
}
/** /**
* Like {@link #getSubCells()} but with the results filtered by a shape. If * Like {@link #getSubCells()} but with the results filtered by a shape. If
@ -196,8 +190,6 @@ public abstract class Cell implements Comparable<Cell> {
*/ */
public abstract Cell getSubCell(Point p); public abstract Cell getSubCell(Point p);
//TODO Cell getSubCell(byte b)
/** /**
* Gets the cells at the next grid cell level that cover this cell. * Gets the cells at the next grid cell level that cover this cell.
* Precondition: Never called when getLevel() == maxLevel. * Precondition: Never called when getLevel() == maxLevel.
@ -211,30 +203,45 @@ public abstract class Cell implements Comparable<Cell> {
*/ */
public abstract int getSubCellsSize(); public abstract int getSubCellsSize();
/** Gets the shape for this cell; typically a Rectangle. This method also serves to trigger any lazy
* loading needed to make the cell instance thread-safe.
*/
public abstract Shape getShape(); public abstract Shape getShape();
/** TODO remove once no longer used. */
public Point getCenter() { public Point getCenter() {
return getShape().getCenter(); return getShape().getCenter();
} }
@Override
public int compareTo(Cell o) {
return getTokenString().compareTo(o.getTokenString());
}
@Override @Override
public boolean equals(Object obj) { public boolean equals(Object obj) {
return !(obj == null || !(obj instanceof Cell)) && getTokenString().equals(((Cell) obj).getTokenString()); //this method isn't "normally" called; just in asserts/tests
if (obj instanceof Cell) {
Cell cell = (Cell) obj;
return getTokenBytes(null).equals(cell.getTokenBytes(null));
} else {
return false;
}
} }
@Override @Override
public int hashCode() { public int hashCode() {
return getTokenString().hashCode(); return getTokenBytesNoLeaf(null).hashCode();
} }
@Override @Override
public String toString() { public String toString() {
return getTokenString() + (isLeaf() ? (char) LEAF_BYTE : ""); //this method isn't "normally" called; just in asserts/tests
return getTokenBytes(null).utf8ToString();
} }
/**
* Returns if the target term is within/underneath this cell; not necessarily a direct descendant.
* @param bytesNoLeaf must be getTokenBytesNoLeaf
* @param term the term
*/
public boolean isWithin(BytesRef bytesNoLeaf, BytesRef term) {
assert bytesNoLeaf.equals(getTokenBytesNoLeaf(null));
return StringHelper.startsWith(term, bytesNoLeaf);
}
} }

View File

@ -83,28 +83,43 @@ public class GeohashPrefixTree extends SpatialPrefixTree {
return new GhCell(GeohashUtils.encodeLatLon(p.getY(), p.getX(), level));//args are lat,lon (y,x) return new GhCell(GeohashUtils.encodeLatLon(p.getY(), p.getX(), level));//args are lat,lon (y,x)
} }
@Override
public Cell getCell(String token) {
return new GhCell(token);
}
@Override @Override
public Cell getCell(byte[] bytes, int offset, int len) { public Cell getCell(byte[] bytes, int offset, int len) {
return new GhCell(bytes, offset, len); return new GhCell(bytes, offset, len);
} }
private static byte[] stringToBytesPlus1(String token) {
//copy ASCII token to byte array with one extra spot for eventual LEAF_BYTE if needed
byte[] bytes = new byte[token.length() + 1];
for (int i = 0; i < token.length(); i++) {
bytes[i] = (byte) token.charAt(i);
}
return bytes;
}
class GhCell extends Cell { class GhCell extends Cell {
GhCell(String token) {
super(token); private Shape shape;//cache
private String geohash;//cache; never has leaf byte, simply a geohash
GhCell(String geohash) {
super(stringToBytesPlus1(geohash), 0, geohash.length());
this.geohash = geohash;
if (isLeaf())
this.geohash = geohash.substring(0, geohash.length() - 1);
} }
GhCell(byte[] bytes, int off, int len) { GhCell(byte[] bytes, int off, int len) {
super(bytes, off, len); super(bytes, off, len);
} }
@Override
protected SpatialPrefixTree getGrid() { return GeohashPrefixTree.this; }
@Override @Override
public void reset(byte[] bytes, int off, int len) { public void reset(byte[] bytes, int off, int len) {
super.reset(bytes, off, len); super.reset(bytes, off, len);
geohash = null;
shape = null; shape = null;
} }
@ -125,26 +140,26 @@ public class GeohashPrefixTree extends SpatialPrefixTree {
@Override @Override
public Cell getSubCell(Point p) { public Cell getSubCell(Point p) {
return GeohashPrefixTree.this.getCell(p, getLevel() + 1);//not performant! return getGrid().getCell(p, getLevel() + 1);//not performant!
} }
private Shape shape;//cache
@Override @Override
public Shape getShape() { public Shape getShape() {
if (shape == null) { if (shape == null) {
shape = GeohashUtils.decodeBoundary(getGeohash(), ctx); shape = GeohashUtils.decodeBoundary(getGeohash(), getGrid().getSpatialContext());
} }
return shape; return shape;
} }
@Override @Override
public Point getCenter() { public Point getCenter() {
return GeohashUtils.decode(getGeohash(), ctx); return GeohashUtils.decode(getGeohash(), getGrid().getSpatialContext());
} }
private String getGeohash() { private String getGeohash() {
return getTokenString(); if (geohash == null)
geohash = getTokenBytesNoLeaf(null).utf8ToString();
return geohash;
} }
}//class GhCell }//class GhCell

View File

@ -22,6 +22,7 @@ import com.spatial4j.core.shape.Point;
import com.spatial4j.core.shape.Rectangle; import com.spatial4j.core.shape.Rectangle;
import com.spatial4j.core.shape.Shape; import com.spatial4j.core.shape.Shape;
import com.spatial4j.core.shape.SpatialRelation; import com.spatial4j.core.shape.SpatialRelation;
import org.apache.lucene.util.BytesRef;
import java.io.PrintStream; import java.io.PrintStream;
import java.text.NumberFormat; import java.text.NumberFormat;
@ -142,15 +143,10 @@ public class QuadPrefixTree extends SpatialPrefixTree {
@Override @Override
public Cell getCell(Point p, int level) { public Cell getCell(Point p, int level) {
List<Cell> cells = new ArrayList<>(1); List<Cell> cells = new ArrayList<>(1);
build(xmid, ymid, 0, cells, new StringBuilder(), ctx.makePoint(p.getX(),p.getY()), level); build(xmid, ymid, 0, cells, new BytesRef(maxLevels+1), ctx.makePoint(p.getX(),p.getY()), level);
return cells.get(0);//note cells could be longer if p on edge return cells.get(0);//note cells could be longer if p on edge
} }
@Override
public Cell getCell(String token) {
return new QuadCell(token);
}
@Override @Override
public Cell getCell(byte[] bytes, int offset, int len) { public Cell getCell(byte[] bytes, int offset, int len) {
return new QuadCell(bytes, offset, len); return new QuadCell(bytes, offset, len);
@ -161,10 +157,10 @@ public class QuadPrefixTree extends SpatialPrefixTree {
double y, double y,
int level, int level,
List<Cell> matches, List<Cell> matches,
StringBuilder str, BytesRef str,
Shape shape, Shape shape,
int maxLevel) { int maxLevel) {
assert str.length() == level; assert str.length == level;
double w = levelW[level] / 2; double w = levelW[level] / 2;
double h = levelH[level] / 2; double h = levelH[level] / 2;
@ -187,51 +183,51 @@ public class QuadPrefixTree extends SpatialPrefixTree {
double cy, double cy,
int level, int level,
List<Cell> matches, List<Cell> matches,
StringBuilder str, BytesRef str,
Shape shape, Shape shape,
int maxLevel) { int maxLevel) {
assert str.length() == level; assert str.length == level;
assert str.offset == 0;
double w = levelW[level] / 2; double w = levelW[level] / 2;
double h = levelH[level] / 2; double h = levelH[level] / 2;
int strlen = str.length(); int strlen = str.length;
Rectangle rectangle = ctx.makeRectangle(cx - w, cx + w, cy - h, cy + h); Rectangle rectangle = ctx.makeRectangle(cx - w, cx + w, cy - h, cy + h);
SpatialRelation v = shape.relate(rectangle); SpatialRelation v = shape.relate(rectangle);
if (SpatialRelation.CONTAINS == v) { if (SpatialRelation.CONTAINS == v) {
str.append(c); str.bytes[str.length++] = (byte)c;//append
//str.append(SpatialPrefixGrid.COVER); //str.append(SpatialPrefixGrid.COVER);
matches.add(new QuadCell(str.toString(),v.transpose())); matches.add(new QuadCell(BytesRef.deepCopyOf(str), v.transpose()));
} else if (SpatialRelation.DISJOINT == v) { } else if (SpatialRelation.DISJOINT == v) {
// nothing // nothing
} else { // SpatialRelation.WITHIN, SpatialRelation.INTERSECTS } else { // SpatialRelation.WITHIN, SpatialRelation.INTERSECTS
str.append(c); str.bytes[str.length++] = (byte)c;//append
int nextLevel = level+1; int nextLevel = level+1;
if (nextLevel >= maxLevel) { if (nextLevel >= maxLevel) {
//str.append(SpatialPrefixGrid.INTERSECTS); //str.append(SpatialPrefixGrid.INTERSECTS);
matches.add(new QuadCell(str.toString(),v.transpose())); matches.add(new QuadCell(BytesRef.deepCopyOf(str), v.transpose()));
} else { } else {
build(cx, cy, nextLevel, matches, str, shape, maxLevel); build(cx, cy, nextLevel, matches, str, shape, maxLevel);
} }
} }
str.setLength(strlen); str.length = strlen;
} }
class QuadCell extends Cell { class QuadCell extends Cell{
public QuadCell(String token) {
super(token);
}
public QuadCell(String token, SpatialRelation shapeRel) {
super(token);
this.shapeRel = shapeRel;
}
QuadCell(byte[] bytes, int off, int len) { QuadCell(byte[] bytes, int off, int len) {
super(bytes, off, len); super(bytes, off, len);
} }
QuadCell(BytesRef str, SpatialRelation shapeRel) {
this(str.bytes, str.offset, str.length);
this.shapeRel = shapeRel;
}
@Override
protected SpatialPrefixTree getGrid() { return QuadPrefixTree.this; }
@Override @Override
public void reset(byte[] bytes, int off, int len) { public void reset(byte[] bytes, int off, int len) {
super.reset(bytes, off, len); super.reset(bytes, off, len);
@ -240,14 +236,26 @@ public class QuadPrefixTree extends SpatialPrefixTree {
@Override @Override
public Collection<Cell> getSubCells() { public Collection<Cell> getSubCells() {
BytesRef source = getTokenBytesNoLeaf(null);
BytesRef target = new BytesRef();
List<Cell> cells = new ArrayList<>(4); List<Cell> cells = new ArrayList<>(4);
cells.add(new QuadCell(getTokenString()+"A")); cells.add(new QuadCell(concat(source, (byte)'A', target), null));
cells.add(new QuadCell(getTokenString()+"B")); cells.add(new QuadCell(concat(source, (byte)'B', target), null));
cells.add(new QuadCell(getTokenString()+"C")); cells.add(new QuadCell(concat(source, (byte)'C', target), null));
cells.add(new QuadCell(getTokenString()+"D")); cells.add(new QuadCell(concat(source, (byte)'D', target), null));
return cells; return cells;
} }
private BytesRef concat(BytesRef source, byte b, BytesRef target) {
assert target.offset == 0;
target.bytes = new byte[source.length + 2];//+2 for new char + potential leaf
target.length = 0;
target.copyBytes(source);
target.bytes[target.length++] = b;
return target;
}
@Override @Override
public int getSubCellsSize() { public int getSubCellsSize() {
return 4; return 4;
@ -268,27 +276,30 @@ public class QuadPrefixTree extends SpatialPrefixTree {
} }
private Rectangle makeShape() { private Rectangle makeShape() {
String token = getTokenString(); BytesRef token = getTokenBytesNoLeaf(null);
double xmin = QuadPrefixTree.this.xmin; double xmin = QuadPrefixTree.this.xmin;
double ymin = QuadPrefixTree.this.ymin; double ymin = QuadPrefixTree.this.ymin;
for (int i = 0; i < token.length(); i++) { for (int i = 0; i < token.length; i++) {
char c = token.charAt(i); byte c = token.bytes[token.offset + i];
if ('A' == c || 'a' == c) { switch (c) {
case 'A':
ymin += levelH[i]; ymin += levelH[i];
} else if ('B' == c || 'b' == c) { break;
case 'B':
xmin += levelW[i]; xmin += levelW[i];
ymin += levelH[i]; ymin += levelH[i];
} else if ('C' == c || 'c' == c) { break;
// nothing really case 'C':
} break;//nothing really
else if('D' == c || 'd' == c) { case 'D':
xmin += levelW[i]; xmin += levelW[i];
} else { break;
default:
throw new RuntimeException("unexpected char: " + c); throw new RuntimeException("unexpected char: " + c);
} }
} }
int len = token.length(); int len = token.length;
double width, height; double width, height;
if (len > 0) { if (len > 0) {
width = levelW[len-1]; width = levelW[len-1];

View File

@ -21,10 +21,12 @@ import com.spatial4j.core.context.SpatialContext;
import com.spatial4j.core.shape.Point; import com.spatial4j.core.shape.Point;
import com.spatial4j.core.shape.Rectangle; import com.spatial4j.core.shape.Rectangle;
import com.spatial4j.core.shape.Shape; import com.spatial4j.core.shape.Shape;
import org.apache.lucene.util.BytesRef;
import java.nio.charset.Charset; import java.nio.charset.Charset;
import java.nio.charset.StandardCharsets; import java.nio.charset.StandardCharsets;
import java.util.ArrayList; import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collection; import java.util.Collection;
import java.util.Collections; import java.util.Collections;
import java.util.List; import java.util.List;
@ -103,14 +105,14 @@ public abstract class SpatialPrefixTree {
private transient Cell worldCell;//cached private transient Cell worldCell;//cached
/** /**
* Returns the level 0 cell which encompasses all spatial data. Equivalent to {@link #getCell(String)} with "". * Returns the level 0 cell which encompasses all spatial data. Equivalent to {@link #getCell(byte[], int, int)} with
* This cell is threadsafe, just like a spatial prefix grid is, although cells aren't * no bytes. This cell is thread-safe, just like a spatial prefix grid is, although cells aren't
* generally threadsafe. * generally thread-safe.
* TODO rename to getTopCell or is this fine?
*/ */
public Cell getWorldCell() { public Cell getWorldCell() {//another possible name: getTopCell
if (worldCell == null) { if (worldCell == null) {
worldCell = getCell(""); worldCell = getCell(BytesRef.EMPTY_BYTES, 0, 0);
worldCell.getShape();//lazy load; make thread-safe
} }
return worldCell; return worldCell;
} }
@ -119,8 +121,6 @@ public abstract class SpatialPrefixTree {
* The cell for the specified token. The empty string should be equal to {@link #getWorldCell()}. * The cell for the specified token. The empty string should be equal to {@link #getWorldCell()}.
* Precondition: Never called when token length > maxLevel. * Precondition: Never called when token length > maxLevel.
*/ */
public abstract Cell getCell(String token);
public abstract Cell getCell(byte[] bytes, int offset, int len); public abstract Cell getCell(byte[] bytes, int offset, int len);
public final Cell getCell(byte[] bytes, int offset, int len, Cell target) { public final Cell getCell(byte[] bytes, int offset, int len, Cell target) {
@ -215,40 +215,23 @@ public abstract class SpatialPrefixTree {
* A Point-optimized implementation of * A Point-optimized implementation of
* {@link #getCells(com.spatial4j.core.shape.Shape, int, boolean, boolean)}. That * {@link #getCells(com.spatial4j.core.shape.Shape, int, boolean, boolean)}. That
* method in facts calls this for points. * method in facts calls this for points.
* <p/>
* This implementation depends on {@link #getCell(String)} being fast, as its
* called repeatedly when incPlarents is true.
*/ */
public List<Cell> getCells(Point p, int detailLevel, boolean inclParents) { public List<Cell> getCells(Point p, int detailLevel, boolean inclParents) {
Cell cell = getCell(p, detailLevel); Cell cell = getCell(p, detailLevel);
if (!inclParents) { assert !cell.isLeaf();
if (!inclParents || detailLevel == 1) {
return Collections.singletonList(cell); return Collections.singletonList(cell);
} }
String endToken = cell.getTokenString(); //fill in reverse order to be sorted
assert endToken.length() == detailLevel; Cell[] cells = new Cell[detailLevel];
List<Cell> cells = new ArrayList<>(detailLevel); for (int i = detailLevel-1; true; i--) {
for (int i = 1; i < detailLevel; i++) { cells[i] = cell;
cells.add(getCell(endToken.substring(0, i)));//TODO refactor: add a cell.getParent() if (i == 0)
break;
cell = cell.getParent();
} }
cells.add(cell); return Arrays.asList(cells);
return cells;
} }
/**
* Will add the trailing leaf byte for leaves. This isn't particularly efficient.
* @deprecated TODO remove; not used and not interesting, don't need collection in & out
*/
public static List<String> cellsToTokenStrings(Collection<Cell> cells) {
List<String> tokens = new ArrayList<>((cells.size()));
for (Cell cell : cells) {
final String token = cell.getTokenString();
if (cell.isLeaf()) {
tokens.add(token + (char) Cell.LEAF_BYTE);
} else {
tokens.add(token);
}
}
return tokens;
}
} }

View File

@ -35,6 +35,9 @@ import org.apache.lucene.spatial.query.SpatialOperation;
import org.junit.Before; import org.junit.Before;
import org.junit.Test; import org.junit.Test;
import java.util.ArrayList;
import java.util.List;
public class SpatialPrefixTreeTest extends SpatialTestCase { public class SpatialPrefixTreeTest extends SpatialTestCase {
//TODO plug in others and test them //TODO plug in others and test them
@ -56,9 +59,10 @@ public class SpatialPrefixTreeTest extends SpatialTestCase {
Cell c = trie.getWorldCell(); Cell c = trie.getWorldCell();
assertEquals(0, c.getLevel()); assertEquals(0, c.getLevel());
assertEquals(ctx.getWorldBounds(), c.getShape()); assertEquals(ctx.getWorldBounds(), c.getShape());
while(c.getLevel() < trie.getMaxLevels()) { while (c.getLevel() < trie.getMaxLevels()) {
prevC = c; prevC = c;
c = c.getSubCells().iterator().next();//TODO random which one? List<Cell> subCells = new ArrayList<>(c.getSubCells());
c = subCells.get(random().nextInt(subCells.size()-1));
assertEquals(prevC.getLevel()+1,c.getLevel()); assertEquals(prevC.getLevel()+1,c.getLevel());
Rectangle prevNShape = (Rectangle) prevC.getShape(); Rectangle prevNShape = (Rectangle) prevC.getShape();