package org.apache.lucene.util;
/**
* 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.NumericTokenStream; // for javadocs
import org.apache.lucene.document.NumericField; // for javadocs
import org.apache.lucene.search.NumericRangeQuery; // for javadocs
import org.apache.lucene.search.NumericRangeFilter; // for javadocs
// TODO: Remove the commented out methods before release!
/**
* This is a helper class to generate prefix-encoded representations for numerical values
* and supplies converters to represent float/double values as sortable integers/longs.
*
* <p>To quickly execute range queries in Apache Lucene, a range is divided recursively
* into multiple intervals for searching: The center of the range is searched only with
* the lowest possible precision in the trie, while the boundaries are matched
* more exactly. This reduces the number of terms dramatically.
*
* <p>This class generates terms to achieve this: First the numerical integer values need to
* be converted to bytes. For that integer values (32 bit or 64 bit) are made unsigned
* and the bits are converted to ASCII chars with each 7 bit. The resulting byte[] is
* sortable like the original integer value (even using UTF-8 sort order). Each value is also
* prefixed (in the first char) by the <code>shift</code> value (number of bits removed) used
* during encoding.
*
* <p>To also index floating point numbers, this class supplies two methods to convert them
* to integer values by changing their bit layout: {@link #doubleToSortableLong},
* {@link #floatToSortableInt}. You will have no precision loss by
* converting floating point numbers to integers and back (only that the integer form
* is not usable). Other data types like dates can easily converted to longs or ints (e.g.
* date to long: {@link java.util.Date#getTime}).
*
* <p>For easy usage, the trie algorithm is implemented for indexing inside
* {@link NumericTokenStream} that can index <code>int</code>, <code>long</code>,
* <code>float</code>, and <code>double</code>. For querying,
* {@link NumericRangeQuery} and {@link NumericRangeFilter} implement the query part
* for the same data types.
*
* <p>This class can also be used, to generate lexicographically sortable (according to
* {@link BytesRef#getUTF8SortedAsUTF16Comparator()}) representations of numeric data
* types for other usages (e.g. sorting).
*
* @lucene.internal
* @since 2.9, API changed non backwards-compliant in 4.0
*/
public final class NumericUtils {
private NumericUtils() {} // no instance!
/**
* The default precision step used by {@link NumericField}, {@link NumericTokenStream},
* {@link NumericRangeQuery}, and {@link NumericRangeFilter} as default
*/
public static final int PRECISION_STEP_DEFAULT = 4;
/**
* Longs are stored at lower precision by shifting off lower bits. The shift count is
* stored as <code>SHIFT_START_LONG+shift</code> in the first byte
*/
public static final byte SHIFT_START_LONG = 0x20;
/**
* The maximum term length (used for <code>byte[]</code> buffer size)
* for encoding <code>long</code> values.
* @see #longToPrefixCoded(long,int,BytesRef)
*/
public static final int BUF_SIZE_LONG = 63/7 + 2;
/**
* Integers are stored at lower precision by shifting off lower bits. The shift count is
* stored as <code>SHIFT_START_INT+shift</code> in the first byte
*/
public static final byte SHIFT_START_INT = 0x60;
/**
* The maximum term length (used for <code>byte[]</code> buffer size)
* for encoding <code>int</code> values.
* @see #intToPrefixCoded(int,int,BytesRef)
*/
public static final int BUF_SIZE_INT = 31/7 + 2;
/**
* Returns prefix coded bits after reducing the precision by <code>shift</code> bits.
* This is method is used by {@link NumericTokenStream}.
* After encoding, {@code bytes.offset} will always be 0.
* @param val the numeric value
* @param shift how many bits to strip from the right
* @param bytes will contain the encoded value
* @return the hash code for indexing (TermsHash)
*/
public static int longToPrefixCoded(final long val, final int shift, final BytesRef bytes) {
if (shift>63 || shift<0)
throw new IllegalArgumentException("Illegal shift value, must be 0..63");
int hash, nChars = (63-shift)/7 + 1;
bytes.offset = 0;
bytes.length = nChars+1;
if (bytes.bytes.length < bytes.length) {
bytes.grow(NumericUtils.BUF_SIZE_LONG);
}
bytes.bytes[0] = (byte) (hash = (SHIFT_START_LONG + shift));
long sortableBits = val ^ 0x8000000000000000L;
sortableBits >>>= shift;
while (nChars > 0) {
// Store 7 bits per byte for compatibility
// with UTF-8 encoding of terms
bytes.bytes[nChars--] = (byte)(sortableBits & 0x7f);
sortableBits >>>= 7;
}
// calculate hash
for (int i = 1; i < bytes.length; i++) {
hash = 31*hash + bytes.bytes[i];
}
return hash;
}
/*
* Returns prefix coded bits after reducing the precision by <code>shift</code> bits.
* This is method is used by {@link LongRangeBuilder}.
* @param val the numeric value
* @param shift how many bits to strip from the right
* @deprecated This method is no longer needed!
*
@Deprecated
public static String longToPrefixCoded(final long val, final int shift) {
final BytesRef buffer = new BytesRef(BUF_SIZE_LONG);
longToPrefixCoded(val, shift, buffer);
return buffer.utf8ToString();
}*/
/*
* This is a convenience method, that returns prefix coded bits of a long without
* reducing the precision. It can be used to store the full precision value as a
* stored field in index.
* <p>To decode, use {@link #prefixCodedToLong}.
* @deprecated This method is no longer needed!
*
@Deprecated
public static String longToPrefixCoded(final long val) {
return longToPrefixCoded(val, 0);
}*/
/**
* Returns prefix coded bits after reducing the precision by <code>shift</code> bits.
* This is method is used by {@link NumericTokenStream}.
* After encoding, {@code bytes.offset} will always be 0.
* @param val the numeric value
* @param shift how many bits to strip from the right
* @param bytes will contain the encoded value
* @return the hash code for indexing (TermsHash)
*/
public static int intToPrefixCoded(final int val, final int shift, final BytesRef bytes) {
if (shift>31 || shift<0)
throw new IllegalArgumentException("Illegal shift value, must be 0..31");
int hash, nChars = (31-shift)/7 + 1;
bytes.offset = 0;
bytes.length = nChars+1;
if (bytes.bytes.length < bytes.length) {
bytes.grow(NumericUtils.BUF_SIZE_INT);
}
bytes.bytes[0] = (byte) (hash = (SHIFT_START_INT + shift));
int sortableBits = val ^ 0x80000000;
sortableBits >>>= shift;
while (nChars > 0) {
// Store 7 bits per byte for compatibility
// with UTF-8 encoding of terms
bytes.bytes[nChars--] = (byte)(sortableBits & 0x7f);
sortableBits >>>= 7;
}
// calculate hash
for (int i = 1; i < bytes.length; i++) {
hash = 31*hash + bytes.bytes[i];
}
return hash;
}
/*
* Returns prefix coded bits after reducing the precision by <code>shift</code> bits.
* This is method is used by {@link IntRangeBuilder}.
* @param val the numeric value
* @param shift how many bits to strip from the right
* @deprecated This method is no longer needed!
*
@Deprecated
public static String intToPrefixCoded(final int val, final int shift) {
final BytesRef buffer = new BytesRef(BUF_SIZE_INT);
intToPrefixCoded(val, shift, buffer);
return buffer.utf8ToString();
}*/
/*
* This is a convenience method, that returns prefix coded bits of an int without
* reducing the precision. It can be used to store the full precision value as a
* stored field in index.
* <p>To decode, use {@link #prefixCodedToInt}.
* @deprecated This method is no longer needed!
*
@Deprecated
public static String intToPrefixCoded(final int val) {
return intToPrefixCoded(val, 0);
}*/
/*
* Returns a long from prefixCoded characters.
* Rightmost bits will be zero for lower precision codes.
* This method can be used to decode e.g. a stored field.
* @throws NumberFormatException if the supplied string is
* not correctly prefix encoded.
* @see #longToPrefixCoded(long)
* @deprecated This method is no longer needed!
*
@Deprecated
public static long prefixCodedToLong(final String prefixCoded) {
return prefixCodedToLong(new BytesRef(prefixCoded));
}*/
/**
* Returns the shift value from a prefix encoded {@code long}.
* @throws NumberFormatException if the supplied {@link BytesRef} is
* not correctly prefix encoded.
*/
public static int getPrefixCodedLongShift(final BytesRef val) {
final int shift = val.bytes[val.offset] - SHIFT_START_LONG;
if (shift > 63 || shift < 0)
throw new NumberFormatException("Invalid shift value in prefixCoded bytes (is encoded value really an INT?)");
return shift;
}
/**
* Returns the shift value from a prefix encoded {@code int}.
* @throws NumberFormatException if the supplied {@link BytesRef} is
* not correctly prefix encoded.
*/
public static int getPrefixCodedIntShift(final BytesRef val) {
final int shift = val.bytes[val.offset] - SHIFT_START_INT;
if (shift > 31 || shift < 0)
throw new NumberFormatException("Invalid shift value in prefixCoded bytes (is encoded value really an INT?)");
return shift;
}
/**
* Returns a long from prefixCoded bytes.
* Rightmost bits will be zero for lower precision codes.
* This method can be used to decode a term's value.
* @throws NumberFormatException if the supplied {@link BytesRef} is
* not correctly prefix encoded.
* @see #longToPrefixCoded(long,int,BytesRef)
*/
public static long prefixCodedToLong(final BytesRef val) {
long sortableBits = 0L;
for (int i=val.offset+1, limit=val.offset+val.length; i<limit; i++) {
sortableBits <<= 7;
final byte b = val.bytes[i];
if (b < 0) {
throw new NumberFormatException(
"Invalid prefixCoded numerical value representation (byte "+
Integer.toHexString(b&0xff)+" at position "+(i-val.offset)+" is invalid)"
);
}
sortableBits |= b;
}
return (sortableBits << getPrefixCodedLongShift(val)) ^ 0x8000000000000000L;
}
/*
* Returns an int from prefixCoded characters.
* Rightmost bits will be zero for lower precision codes.
* This method can be used to decode a term's value.
* @throws NumberFormatException if the supplied string is
* not correctly prefix encoded.
* @see #intToPrefixCoded(int)
* @deprecated This method is no longer needed!
*
@Deprecated
public static int prefixCodedToInt(final String prefixCoded) {
return prefixCodedToInt(new BytesRef(prefixCoded));
}*/
/*
* Returns an int from prefixCoded bytes.
* Rightmost bits will be zero for lower precision codes.
* This method can be used to decode a term's value.
* @throws NumberFormatException if the supplied {@link BytesRef} is
* not correctly prefix encoded.
* @see #intToPrefixCoded(int,int,BytesRef)
*/
public static int prefixCodedToInt(final BytesRef val) {
int sortableBits = 0;
for (int i=val.offset+1, limit=val.offset+val.length; i<limit; i++) {
sortableBits <<= 7;
final byte b = val.bytes[i];
if (b < 0) {
throw new NumberFormatException(
"Invalid prefixCoded numerical value representation (byte "+
Integer.toHexString(b&0xff)+" at position "+(i-val.offset)+" is invalid)"
);
}
sortableBits |= b;
}
return (sortableBits << getPrefixCodedIntShift(val)) ^ 0x80000000;
}
/**
* Converts a <code>double</code> value to a sortable signed <code>long</code>.
* The value is converted by getting their IEEE 754 floating-point "double format"
* bit layout and then some bits are swapped, to be able to compare the result as long.
* By this the precision is not reduced, but the value can easily used as a long.
* @see #sortableLongToDouble
*/
public static long doubleToSortableLong(double val) {
long f = Double.doubleToRawLongBits(val);
if (f<0) f ^= 0x7fffffffffffffffL;
return f;
}
/*
* Convenience method: this just returns:
* longToPrefixCoded(doubleToSortableLong(val))
* @deprecated This method is no longer needed!
*
@Deprecated
public static String doubleToPrefixCoded(double val) {
return longToPrefixCoded(doubleToSortableLong(val));
}*/
/**
* Converts a sortable <code>long</code> back to a <code>double</code>.
* @see #doubleToSortableLong
*/
public static double sortableLongToDouble(long val) {
if (val<0) val ^= 0x7fffffffffffffffL;
return Double.longBitsToDouble(val);
}
/*
* Convenience method: this just returns:
* sortableLongToDouble(prefixCodedToLong(val))
* @deprecated This method is no longer needed!
*
@Deprecated
public static double prefixCodedToDouble(String val) {
return sortableLongToDouble(prefixCodedToLong(val));
}*/
/**
* Converts a <code>float</code> value to a sortable signed <code>int</code>.
* The value is converted by getting their IEEE 754 floating-point "float format"
* bit layout and then some bits are swapped, to be able to compare the result as int.
* By this the precision is not reduced, but the value can easily used as an int.
* @see #sortableIntToFloat
*/
public static int floatToSortableInt(float val) {
int f = Float.floatToRawIntBits(val);
if (f<0) f ^= 0x7fffffff;
return f;
}
/*
* Convenience method: this just returns:
* intToPrefixCoded(floatToSortableInt(val))
* @deprecated This method is no longer needed!
*
@Deprecated
public static String floatToPrefixCoded(float val) {
return intToPrefixCoded(floatToSortableInt(val));
}*/
/**
* Converts a sortable <code>int</code> back to a <code>float</code>.
* @see #floatToSortableInt
*/
public static float sortableIntToFloat(int val) {
if (val<0) val ^= 0x7fffffff;
return Float.intBitsToFloat(val);
}
/*
* Convenience method: this just returns:
* sortableIntToFloat(prefixCodedToInt(val))
* @deprecated This method is no longer needed!
*
@Deprecated
public static float prefixCodedToFloat(String val) {
return sortableIntToFloat(prefixCodedToInt(val));
}*/
/**
* Splits a long range recursively.
* You may implement a builder that adds clauses to a
* {@link org.apache.lucene.search.BooleanQuery} for each call to its
* {@link LongRangeBuilder#addRange(BytesRef,BytesRef)}
* method.
* <p>This method is used by {@link NumericRangeQuery}.
*/
public static void splitLongRange(final LongRangeBuilder builder,
final int precisionStep, final long minBound, final long maxBound
) {
splitRange(builder, 64, precisionStep, minBound, maxBound);
}
/**
* Splits an int range recursively.
* You may implement a builder that adds clauses to a
* {@link org.apache.lucene.search.BooleanQuery} for each call to its
* {@link IntRangeBuilder#addRange(BytesRef,BytesRef)}
* method.
* <p>This method is used by {@link NumericRangeQuery}.
*/
public static void splitIntRange(final IntRangeBuilder builder,
final int precisionStep, final int minBound, final int maxBound
) {
splitRange(builder, 32, precisionStep, minBound, maxBound);
}
/** This helper does the splitting for both 32 and 64 bit. */
private static void splitRange(
final Object builder, final int valSize,
final int precisionStep, long minBound, long maxBound
) {
if (precisionStep < 1)
throw new IllegalArgumentException("precisionStep must be >=1");
if (minBound > maxBound) return;
for (int shift=0; ; shift += precisionStep) {
// calculate new bounds for inner precision
final long diff = 1L << (shift+precisionStep),
mask = ((1L<<precisionStep) - 1L) << shift;
final boolean
hasLower = (minBound & mask) != 0L,
hasUpper = (maxBound & mask) != mask;
final long
nextMinBound = (hasLower ? (minBound + diff) : minBound) & ~mask,
nextMaxBound = (hasUpper ? (maxBound - diff) : maxBound) & ~mask;
final boolean
lowerWrapped = nextMinBound < minBound,
upperWrapped = nextMaxBound > maxBound;
if (shift+precisionStep>=valSize || nextMinBound>nextMaxBound || lowerWrapped || upperWrapped) {
// We are in the lowest precision or the next precision is not available.
addRange(builder, valSize, minBound, maxBound, shift);
// exit the split recursion loop
break;
}
if (hasLower)
addRange(builder, valSize, minBound, minBound | mask, shift);
if (hasUpper)
addRange(builder, valSize, maxBound & ~mask, maxBound, shift);
// recurse to next precision
minBound = nextMinBound;
maxBound = nextMaxBound;
}
}
/** Helper that delegates to correct range builder */
private static void addRange(
final Object builder, final int valSize,
long minBound, long maxBound,
final int shift
) {
// for the max bound set all lower bits (that were shifted away):
// this is important for testing or other usages of the splitted range
// (e.g. to reconstruct the full range). The prefixEncoding will remove
// the bits anyway, so they do not hurt!
maxBound |= (1L << shift) - 1L;
// delegate to correct range builder
switch(valSize) {
case 64:
((LongRangeBuilder)builder).addRange(minBound, maxBound, shift);
break;
case 32:
((IntRangeBuilder)builder).addRange((int)minBound, (int)maxBound, shift);
break;
default:
// Should not happen!
throw new IllegalArgumentException("valSize must be 32 or 64.");
}
}
/**
* Callback for {@link #splitLongRange}.
* You need to overwrite only one of the methods.
* @lucene.internal
* @since 2.9, API changed non backwards-compliant in 4.0
*/
public static abstract class LongRangeBuilder {
/**
* Overwrite this method, if you like to receive the already prefix encoded range bounds.
* You can directly build classical (inclusive) range queries from them.
*/
public void addRange(BytesRef minPrefixCoded, BytesRef maxPrefixCoded) {
throw new UnsupportedOperationException();
}
/**
* Overwrite this method, if you like to receive the raw long range bounds.
* You can use this for e.g. debugging purposes (print out range bounds).
*/
public void addRange(final long min, final long max, final int shift) {
final BytesRef minBytes = new BytesRef(BUF_SIZE_LONG), maxBytes = new BytesRef(BUF_SIZE_LONG);
longToPrefixCoded(min, shift, minBytes);
longToPrefixCoded(max, shift, maxBytes);
addRange(minBytes, maxBytes);
}
}
/**
* Callback for {@link #splitIntRange}.
* You need to overwrite only one of the methods.
* @lucene.internal
* @since 2.9, API changed non backwards-compliant in 4.0
*/
public static abstract class IntRangeBuilder {
/**
* Overwrite this method, if you like to receive the already prefix encoded range bounds.
* You can directly build classical range (inclusive) queries from them.
*/
public void addRange(BytesRef minPrefixCoded, BytesRef maxPrefixCoded) {
throw new UnsupportedOperationException();
}
/**
* Overwrite this method, if you like to receive the raw int range bounds.
* You can use this for e.g. debugging purposes (print out range bounds).
*/
public void addRange(final int min, final int max, final int shift) {
final BytesRef minBytes = new BytesRef(BUF_SIZE_INT), maxBytes = new BytesRef(BUF_SIZE_INT);
intToPrefixCoded(min, shift, minBytes);
intToPrefixCoded(max, shift, maxBytes);
addRange(minBytes, maxBytes);
}
}
}