src.it.unimi.dsi.bits.TransformationStrategies Maven / Gradle / Ivy
Show all versions of dsiutils Show documentation
/*
* DSI utilities
*
* Copyright (C) 2007-2020 Sebastiano Vigna
*
* This library is free software; you can redistribute it and/or modify it
* under the terms of the GNU Lesser General Public License as published by the Free
* Software Foundation; either version 3 of the License, or (at your option)
* any later version.
*
* This library is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
* or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License
* for more details.
*
* You should have received a copy of the GNU Lesser General Public License
* along with this program; if not, see This class provides several {@linkplain TransformationStrategy transformation strategies} that turn
* strings or other objects into bit vectors. The transformations might optionally be:
*
* - Lexicographical: for objects based on bytes or characters, such as strings
* and byte arrays, this means that the first bit of the bit vector is the most significant
* bit of the first byte or character, and so on. In other word, the lexicographical order between
* bit vectors reflects the lexicographical byte-by-byte, char-by-char, etc. order. Thiss property
* is necessary for some kind of static structure that depends on it, but it has some computational
* cost, as after compacting byte or chars into a long we need to revert the bit order of each piece.
*
- Prefix-free: no two bit vector returned by the transformation on two
* different objects will be comparable in prefix order. Again, this might require to use more
* linear (e.g., {@link #prefixFree()}) or constant (e.g., {@link #prefixFreeIso()}) additional space.
*
*
* As a general rule, transformations without additional naming are lexicographical.
* Transformation that generate prefix-free bit vectors are marked as such.
* Plain transformations that do not provide any guarantee are called raw. They should be
* used only when performance is the main issue and the two properties above are not relevant.
*
* @see TransformationStrategy
*/
@SuppressWarnings("fallthrough")
public class TransformationStrategies {
private final static TransformationStrategy IDENTITY = new TransformationStrategy() {
private static final long serialVersionUID = 1L;
@Override
public BitVector toBitVector(final BitVector v) {
return v;
}
@Override
public long length(final BitVector v) {
return v.length();
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
public Object readResolve() {
return IDENTITY;
}
};
/** Reverses the bit order in the bytes of the provided word.
*
* @param word a word.
* @return {@code word}, with the bit order inside each byte reversed.
*/
private static final long reverseBytes(long word) {
word = (word & 0x5555555555555555L) << 1 | (word >>> 1) & 0x5555555555555555L;
word = (word & 0x3333333333333333L) << 2 | (word >>> 2) & 0x3333333333333333L;
return (word & 0x0f0f0f0f0f0f0f0fL) << 4 | (word >>> 4) & 0x0f0f0f0f0f0f0f0fL;
}
/** Reverses the bit order in the characters of the provided word.
*
* @param word a word.
* @return {@code word}, with the bit order inside each 16-bit character reversed.
*/
private static final long reverseChars(long word) {
word = (word & 0x5555555555555555L) << 1 | (word >>> 1) & 0x5555555555555555L;
word = (word & 0x3333333333333333L) << 2 | (word >>> 2) & 0x3333333333333333L;
word = (word & 0x0f0f0f0f0f0f0f0fL) << 4 | (word >>> 4) & 0x0f0f0f0f0f0f0f0fL;
return (word & 0x00ff00ff00ff00ffL) << 8 | (word >>> 8) & 0x00ff00ff00ff00ffL;
}
/** A trivial transformation for data already in {@link BitVector} form. */
@SuppressWarnings("unchecked")
public static TransformationStrategy identity() {
return (TransformationStrategy)IDENTITY;
}
private static final TransformationStrategy RAW_UTF32 = new RawUtf32TransformationStrategy();
/**A trivial raw transformation from strings to bit vectors
* that turns the UTF-16 representation into a UTF-32 representation,
* decodes surrogate pairs and concatenates the bits of the UTF-32 representation.
*
* Warning: this transformation is not lexicographic.
*/
@SuppressWarnings("unchecked")
public static TransformationStrategy rawUtf32() {
return (TransformationStrategy)RAW_UTF32;
}
private static class RawUtf32TransformationStrategy implements TransformationStrategy, Serializable {
private static final long serialVersionUID = 1L;
@Override
public long length(final CharSequence cs) {
return length(cs.toString());
}
private long length(final String s) {
return s.codePointCount(0, s.length()) * (long)Integer.SIZE;
}
@Override
public BitVector toBitVector(final CharSequence cs) {
final String s = cs.toString();
final int length = s.length();
final LongArrayBitVector bitVector = LongArrayBitVector.getInstance(length(s));
for (int i = 0, cp; i < length; i += Character.charCount(cp))
bitVector.append(cp = s.codePointAt(i), Integer.SIZE);
return bitVector;
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
private Object readResolve() {
return RAW_UTF32;
}
}
private static final TransformationStrategy UTF32 = new Utf32TransformationStrategy(false);
/** A transformation from strings to bit vectors that turns the UTF-16 representation into a UTF-32 representation,
* decodes surrogate pairs and concatenates the bits of the UTF-32 representation.
*/
@SuppressWarnings("unchecked")
public static TransformationStrategy utf32() {
return (TransformationStrategy)UTF32;
}
private static final TransformationStrategy PREFIX_FREE_UTF32 = new Utf32TransformationStrategy(true);
/** A transformation from strings to bit vectors that turns the UTF-16 representation into a UTF-32 representation,
* decodes surrogate pairs, concatenates the bits of the UTF-32 representation and completes
* the representation with an NUL to guarantee lexicographical ordering and prefix-freeness.
*
* Note that strings provided to this strategy must not contain NULs.
*/
@SuppressWarnings("unchecked")
public static TransformationStrategy prefixFreeUtf32() {
return (TransformationStrategy)PREFIX_FREE_UTF32;
}
private static class Utf32TransformationStrategy implements TransformationStrategy, Serializable {
private static final long serialVersionUID = 1L;
/** Whether we should guarantee prefix-freeness by adding 0 to the end of each string. */
private final boolean prefixFree;
/** Creates a UTF32 transformation strategy. The strategy will map a string to its UTF32 bit sequence by decoding surrogate pairs.
*
* @param prefixFree if true, the resulting set of binary words will be made prefix free by adding a NUL at the end of the string.
*/
protected Utf32TransformationStrategy(final boolean prefixFree) {
this.prefixFree = prefixFree;
}
@Override
public long length(final CharSequence cs) {
return length(cs.toString());
}
private long length(final String s) {
return (s.codePointCount(0, s.length()) + (prefixFree ? 1 : 0)) * (long)Integer.SIZE;
}
@Override
public BitVector toBitVector(final CharSequence cs) {
final String s = cs.toString();
final int length = s.length();
final LongArrayBitVector bitVector = LongArrayBitVector.getInstance(length(s));
for (int i = 0, cp; i < length; i += Character.charCount(cp))
bitVector.append(Integer.reverse(cp = s.codePointAt(i)) & -1L >>> 32, Integer.SIZE);
if (prefixFree) bitVector.append(0, Integer.SIZE);
return bitVector;
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
private Object readResolve() {
return prefixFree ? PREFIX_FREE_UTF32 : UTF32;
}
}
private static final TransformationStrategy RAW_UTF16 = new RawUtf16TransformationStrategy();
/** A trivial, high-performance, raw transformation from strings to bit vectors that concatenates the bits of the UTF-16 representation.
*
* Warning: this transformation is not lexicographic.
*
*
Warning: bit vectors returned by this strategy are adaptors around the original string. If the string
* changes while the bit vector is being accessed, the results will be unpredictable.
*/
@SuppressWarnings("unchecked")
public static TransformationStrategy rawUtf16() {
return (TransformationStrategy)RAW_UTF16;
}
private static class RawUtf16TransformationStrategy implements TransformationStrategy, Serializable {
private static final long serialVersionUID = 1L;
@Override
public long length(final CharSequence s) {
return s.length() * (long)Character.SIZE;
}
private static class RawUtf16MutableStringBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 1L;
private static final int CHAR_MASK = Character.SIZE - 1;
private static final int LOG2_CHAR_SIZE = 4;
private final char[] a;
private final long length;
public RawUtf16MutableStringBitVector(final MutableString s) {
this.a = s.array();
length = s.length() * (long)Character.SIZE;
}
@Override
public boolean getBoolean(final long index) {
if (index > length) throw new IndexOutOfBoundsException();
final int charIndex = (int)(index >>> LOG2_CHAR_SIZE);
return (a[charIndex] & 1 << (index & CHAR_MASK)) != 0;
}
@Override
public long getLong(final long from, final long to) {
final int startBit = (int)(from & CHAR_MASK);
if (startBit == 0) {
final int pos = (int)(from >>> LOG2_CHAR_SIZE);
if (to == from + Long.SIZE)
return (long)a[pos + 3] << 48 | (long)a[pos + 2] << 32 | (long)a[pos + 1] << 16 | a[pos];
if ((to & CHAR_MASK) == 0) {
long word = 0;
switch ((int)((to - from) >>> LOG2_CHAR_SIZE)) {
case 3: word |= (long)a[pos + 2] << 32;
case 2: word |= (long)a[pos + 1] << 16;
case 1: word |= a[pos + 0];
}
return word;
}
}
final long l = Long.SIZE - (to - from);
final long startPos = from - startBit;
if (l == Long.SIZE) return 0;
if (startBit <= l) return getLong(startPos, Math.min(length, startPos + Long.SIZE)) << l - startBit >>> l;
return getLong(startPos, startPos + Long.SIZE) >>> startBit | getLong(startPos + Long.SIZE, Math.min(length, startPos + 2 * Long.SIZE)) << Long.SIZE + l - startBit >>> l;
}
@Override
public long length() {
return length;
}
}
private static class RawUtf16CharSequenceBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 1L;
private static final int CHAR_MASK = Character.SIZE - 1;
private static final int LOG2_CHAR_SIZE = 4;
private final CharSequence s;
private final long length;
public RawUtf16CharSequenceBitVector(final CharSequence s) {
this.s = s;
length = s.length() * (long)Character.SIZE;
}
@Override
public boolean getBoolean(final long index) {
if (index > length) throw new IndexOutOfBoundsException();
final int charIndex = (int)(index >>> LOG2_CHAR_SIZE);
return (s.charAt(charIndex) & 1 << (index & CHAR_MASK)) != 0;
}
@Override
public long getLong(final long from, final long to) {
final int startBit = (int)(from & CHAR_MASK);
if (startBit == 0) {
final int pos = (int)(from >>> LOG2_CHAR_SIZE);
if (to == from + Long.SIZE)
return (long)s.charAt(pos + 3) << 48 | (long)s.charAt(pos + 2) << 32 | (long)s.charAt(pos + 1) << 16 | s.charAt(pos);
if ((to & CHAR_MASK) == 0) {
long word = 0;
switch ((int)((to - from) >>> LOG2_CHAR_SIZE)) {
case 3: word |= (long)s.charAt(pos + 2) << 32;
case 2: word |= (long)s.charAt(pos + 1) << 16;
case 1: word |= s.charAt(pos + 0);
}
return word;
}
}
final long l = Long.SIZE - (to - from);
final long startPos = from - startBit;
if (l == Long.SIZE) return 0;
if (startBit <= l) return getLong(startPos, Math.min(length, startPos + Long.SIZE)) << l - startBit >>> l;
return getLong(startPos, startPos + Long.SIZE) >>> startBit | getLong(startPos + Long.SIZE, Math.min(length, startPos + 2 * Long.SIZE)) << Long.SIZE + l - startBit >>> l;
}
@Override
public long length() {
return length;
}
}
@Override
public BitVector toBitVector(final CharSequence s) {
return s instanceof MutableString ? new RawUtf16MutableStringBitVector((MutableString)s) : new RawUtf16CharSequenceBitVector(s);
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
private Object readResolve() {
return RAW_UTF16;
}
}
private static final TransformationStrategy UTF16 = new Utf16TransformationStrategy(false);
/** A trivial transformation from strings to bit vectors that concatenates the bits of the UTF-16 representation.
*
* Warning: bit vectors returned by this strategy are adaptors around the original string. If the string
* changes while the bit vector is being accessed, the results will be unpredictable.
*/
@SuppressWarnings("unchecked")
public static TransformationStrategy utf16() {
return (TransformationStrategy)UTF16;
}
private static final TransformationStrategy PREFIX_FREE_UTF16 = new Utf16TransformationStrategy(true);
/** A trivial transformation from strings to bit vectors that concatenates the bits of the UTF-16 representation and completes
* the representation with an NUL to guarantee lexicographical ordering and prefix-freeness.
*
* Note that strings provided to this strategy must not contain NULs.
*
*
Warning: bit vectors returned by this strategy are adaptors around the original string. If the string
* changes while the bit vector is being accessed, the results will be unpredictable.
*/
@SuppressWarnings("unchecked")
public static TransformationStrategy prefixFreeUtf16() {
return (TransformationStrategy)PREFIX_FREE_UTF16;
}
private static class Utf16TransformationStrategy implements TransformationStrategy, Serializable {
private static final long serialVersionUID = 1L;
/** Whether we should guarantee prefix-freeness by adding 0 to the end of each string. */
private final boolean prefixFree;
/** Creates a UTF16 transformation strategy. The strategy will map a string to its natural UTF16 bit sequence.
*
* @param prefixFree if true, the resulting set of binary words will be made prefix free by adding a NUL at the end of the string.
*/
protected Utf16TransformationStrategy(final boolean prefixFree) {
this.prefixFree = prefixFree;
}
@Override
public long length(final CharSequence s) {
return (s.length() + (prefixFree ? 1 : 0)) * (long)Character.SIZE;
}
private static class Utf16CharSequenceBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 1L;
private static final int CHAR_MASK = Character.SIZE - 1;
private static final int LOG2_CHAR_SIZE = 4;
private final CharSequence s;
private final long length;
private final long actualEnd;
public Utf16CharSequenceBitVector(final CharSequence s, final boolean prefixFree) {
this.s = s;
actualEnd = s.length() * (long)Character.SIZE;
length = actualEnd + (prefixFree ? Character.SIZE : 0);
}
@Override
public boolean getBoolean(final long index) {
if (index > length) throw new IndexOutOfBoundsException();
if (index >= actualEnd) return false;
final int charIndex = (int)(index >>> LOG2_CHAR_SIZE);
return (s.charAt(charIndex) & 0x8000 >>> (index & CHAR_MASK)) != 0;
}
@Override
public long getLong(final long from, final long to) {
final int startBit = (int)(from & CHAR_MASK);
if (startBit == 0) {
final int pos = (int)(from >>> LOG2_CHAR_SIZE);
if (to == from + Long.SIZE)
return reverseChars((to > actualEnd ? 0 : (long)s.charAt(pos + 3)) << 48 | (long)s.charAt(pos + 2) << 32 | (long)s.charAt(pos + 1) << 16 | s.charAt(pos));
if ((to & CHAR_MASK) == 0) {
long word = 0;
switch ((int)((Math.min(to, actualEnd) - Math.min(from, actualEnd)) >>> LOG2_CHAR_SIZE)) {
case 3: word |= (long)s.charAt(pos + 2) << 32;
case 2: word |= (long)s.charAt(pos + 1) << 16;
case 1: word |= s.charAt(pos);
}
return reverseChars(word);
}
}
final long l = Long.SIZE - (to - from);
final long startPos = from - startBit;
if (l == Long.SIZE) return 0;
if (startBit <= l) return getLong(startPos, Math.min(length, startPos + Long.SIZE)) << l - startBit >>> l;
return getLong(startPos, startPos + Long.SIZE) >>> startBit | getLong(startPos + Long.SIZE, Math.min(length, startPos + 2 * Long.SIZE)) << Long.SIZE + l - startBit >>> l;
}
@Override
public long length() {
return length;
}
}
private static class Utf16MutableStringBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 1L;
private static final int CHAR_MASK = Character.SIZE - 1;
private static final int LOG2_CHAR_SIZE = 4;
private final char[] a;
private final long length;
private final long actualEnd;
public Utf16MutableStringBitVector(final MutableString s, final boolean prefixFree) {
this.a = s.array();
actualEnd = s.length() * (long)Character.SIZE;
length = actualEnd + (prefixFree ? Character.SIZE : 0);
}
@Override
public boolean getBoolean(final long index) {
if (index > length) throw new IndexOutOfBoundsException();
if (index >= actualEnd) return false;
final int charIndex = (int)(index >>> LOG2_CHAR_SIZE);
return (a[charIndex] & 0x8000 >>> (index & CHAR_MASK)) != 0;
}
@Override
public long getLong(final long from, final long to) {
final int startBit = (int)(from & CHAR_MASK);
if (startBit == 0) {
final int pos = (int)(from >>> LOG2_CHAR_SIZE);
if (to == from + Long.SIZE)
return reverseChars((to > actualEnd ? 0 : (long)a[pos + 3]) << 48 | (long)a[pos + 2] << 32 | (long)a[pos + 1] << 16 | a[pos]);
if ((to & CHAR_MASK) == 0) {
long word = 0;
switch ((int)((Math.min(to, actualEnd) - Math.min(from, actualEnd)) >>> LOG2_CHAR_SIZE)) {
case 3: word |= (long)a[pos + 2] << 32;
case 2: word |= (long)a[pos + 1] << 16;
case 1: word |= a[pos];
}
return reverseChars(word);
}
}
final long l = Long.SIZE - (to - from);
final long startPos = from - startBit;
if (l == Long.SIZE) return 0;
if (startBit <= l) return getLong(startPos, Math.min(length, startPos + Long.SIZE)) << l - startBit >>> l;
return getLong(startPos, startPos + Long.SIZE) >>> startBit | getLong(startPos + Long.SIZE, Math.min(length, startPos + 2 * Long.SIZE)) << Long.SIZE + l - startBit >>> l;
}
@Override
public long length() {
return length;
}
}
@Override
public BitVector toBitVector(final CharSequence s) {
return s instanceof MutableString ? new Utf16MutableStringBitVector((MutableString)s, prefixFree) : new Utf16CharSequenceBitVector(s, prefixFree);
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
private Object readResolve() {
return prefixFree ? PREFIX_FREE_UTF16 : UTF16;
}
}
private static final TransformationStrategy RAW_ISO = new RawISOTransformationStrategy();
/** A trivial, high-performance, raw transformation from strings to bit vectors that concatenates the lower eight bits bits of the UTF-16 representation.
*
* Warning: this transformation is not lexicographic.
*
*
Note that this transformation is sensible only for strings that are known to be contain just characters in the ISO-8859-1 charset.
*
*
Warning: bit vectors returned by this strategy are adaptors around the original string. If the string
* changes while the bit vector is being accessed, the results will be unpredictable.
*/
@SuppressWarnings("unchecked")
public static TransformationStrategy rawIso() {
return (TransformationStrategy)RAW_ISO;
}
private static class RawISOTransformationStrategy implements TransformationStrategy, Serializable {
private static final long serialVersionUID = 1L;
private static final int LOG2_BYTE_SIZE = 3;
private static final long BYTE_MASK = Byte.SIZE - 1;
@Override
public long length(final CharSequence s) {
return s.length() * (long)Byte.SIZE;
}
private static class RawISOCharSequenceBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 1L;
private final CharSequence s;
private final long length;
public RawISOCharSequenceBitVector(final CharSequence s) {
this.s = s;
length = s.length() * (long)Byte.SIZE;
}
@Override
public boolean getBoolean(final long index) {
if (index > length) throw new IndexOutOfBoundsException();
final int byteIndex = (int)(index >>> LOG2_BYTE_SIZE);
return (s.charAt(byteIndex) & 1 << (index & BYTE_MASK)) != 0;
}
@Override
public long getLong(final long from, final long to) {
final int startBit = (int)(from & BYTE_MASK);
if (startBit == 0) {
final int pos = (int)(from >>> LOG2_BYTE_SIZE);
if (to == from + Long.SIZE)
return (s.charAt(pos + 7) & 0xFFL) << 56 |
(s.charAt(pos + 6) & 0xFFL) << 48 |
(s.charAt(pos + 5) & 0xFFL) << 40 |
(s.charAt(pos + 4) & 0xFFL) << 32 |
(s.charAt(pos + 3) & 0xFFL) << 24 |
(s.charAt(pos + 2) & 0xFF) << 16 |
(s.charAt(pos + 1) & 0xFF) << 8 |
(s.charAt(pos) & 0xFF);
if ((to & BYTE_MASK) == 0) {
long word = 0;
switch ((int)((to - from) >>> LOG2_BYTE_SIZE)) {
case 7: word |= (s.charAt(pos + 6) & 0xFFL) << 48;
case 6: word |= (s.charAt(pos + 5) & 0xFFL) << 40;
case 5: word |= (s.charAt(pos + 4) & 0xFFL) << 32;
case 4: word |= (s.charAt(pos + 3) & 0xFFL) << 24;
case 3: word |= (s.charAt(pos + 2) & 0xFF) << 16;
case 2: word |= (s.charAt(pos + 1) & 0xFF) << 8;
case 1: word |= s.charAt(pos) & 0xFF;
}
return word;
}
}
final long l = Long.SIZE - (to - from);
final long startPos = from - startBit;
if (l == Long.SIZE) return 0;
if (startBit <= l) return getLong(startPos, Math.min(length, startPos + Long.SIZE)) << l - startBit >>> l;
return getLong(startPos, startPos + Long.SIZE) >>> startBit | getLong(startPos + Long.SIZE, Math.min(length, startPos + 2 * Long.SIZE)) << Long.SIZE + l - startBit >>> l;
}
@Override
public long length() {
return length;
}
}
private static class RawISOMutableStringBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 1L;
private final char[] a;
private final long length;
public RawISOMutableStringBitVector(final MutableString s) {
this.a = s.array();
length = s.length() * (long)Byte.SIZE;
}
@Override
public boolean getBoolean(final long index) {
if (index > length) throw new IndexOutOfBoundsException();
final int byteIndex = (int)(index >>> LOG2_BYTE_SIZE);
return (a[byteIndex] & 1 << (index & BYTE_MASK)) != 0;
}
@Override
public long getLong(final long from, final long to) {
final int startBit = (int)(from & BYTE_MASK);
if (startBit == 0) {
final int pos = (int)(from >>> LOG2_BYTE_SIZE);
if (to == from + Long.SIZE)
return (a[pos + 7] & 0xFFL) << 56 |
(a[pos + 6] & 0xFFL) << 48 |
(a[pos + 5] & 0xFFL) << 40 |
(a[pos + 4] & 0xFFL) << 32 |
(a[pos + 3] & 0xFFL) << 24 |
(a[pos + 2] & 0xFF) << 16 |
(a[pos + 1] & 0xFF) << 8 |
(a[pos] & 0xFF);
if ((to & BYTE_MASK) == 0) {
long word = 0;
switch ((int)((to - from) >>> LOG2_BYTE_SIZE)) {
case 7: word |= (a[pos + 6] & 0xFFL) << 48;
case 6: word |= (a[pos + 5] & 0xFFL) << 40;
case 5: word |= (a[pos + 4] & 0xFFL) << 32;
case 4: word |= (a[pos + 3] & 0xFFL) << 24;
case 3: word |= (a[pos + 2] & 0xFF) << 16;
case 2: word |= (a[pos + 1] & 0xFF) << 8;
case 1: word |= a[pos] & 0xFF;
}
return word;
}
}
final long l = Long.SIZE - (to - from);
final long startPos = from - startBit;
if (l == Long.SIZE) return 0;
if (startBit <= l) return getLong(startPos, Math.min(length, startPos + Long.SIZE)) << l - startBit >>> l;
return getLong(startPos, startPos + Long.SIZE) >>> startBit | getLong(startPos + Long.SIZE, Math.min(length, startPos + 2 * Long.SIZE)) << Long.SIZE + l - startBit >>> l;
}
@Override
public long length() {
return length;
}
}
@Override
public BitVector toBitVector(final CharSequence s) {
return s instanceof MutableString ? new RawISOMutableStringBitVector((MutableString)s) : new RawISOCharSequenceBitVector(s);
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
private Object readResolve() {
return RAW_ISO;
}
}
private static final TransformationStrategy ISO = new ISOTransformationStrategy(false);
/** A trivial transformation from strings to bit vectors that concatenates the lower eight bits of the UTF-16 representation.
*
* Note that this transformation is sensible only for strings that are known to be contain just characters in the ISO-8859-1 charset.
*
*
Warning: bit vectors returned by this strategy are adaptors around the original string. If the string
* changes while the bit vector is being accessed, the results will be unpredictable.
*/
@SuppressWarnings("unchecked")
public static TransformationStrategy iso() {
return (TransformationStrategy)ISO;
}
private static final TransformationStrategy PREFIX_FREE_ISO = new ISOTransformationStrategy(true);
/** A trivial transformation from strings to bit vectors that concatenates the lower eight bits bits of the UTF-16 representation and completes
* the representation with an ASCII NUL to guarantee lexicographical ordering and prefix-freeness.
*
* Note that this transformation is sensible only for strings that are known to be contain just characters in the ISO-8859-1 charset, and
* that strings provided to this strategy must not contain ASCII NULs.
*
*
Warning: bit vectors returned by this strategy are adaptors around the original string. If the string
* changes while the bit vector is being accessed, the results will be unpredictable.
*/
@SuppressWarnings("unchecked")
public static TransformationStrategy prefixFreeIso() {
return (TransformationStrategy)PREFIX_FREE_ISO;
}
private static class ISOTransformationStrategy implements TransformationStrategy, Serializable {
private static final long serialVersionUID = 1L;
/** Whether we should guarantee prefix-freeness by adding 0 to the end of each string. */
private final boolean prefixFree;
/** Creates an ISO transformation strategy. The strategy will map a string to the lowest eight bits of its natural UTF16 bit sequence.
*
* @param prefixFree if true, the resulting set of binary words will be made prefix free by adding a NUL at the end of the string.
*/
protected ISOTransformationStrategy(final boolean prefixFree) {
this.prefixFree = prefixFree;
}
@Override
public long length(final CharSequence s) {
return (s.length() + (prefixFree ? 1 : 0)) * (long)Byte.SIZE;
}
private static class ISOCharSequenceBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 1L;
private static final int LOG2_BYTE_SIZE = 3;
private static final long BYTE_MASK = Byte.SIZE - 1;
private final CharSequence s;
private final long length;
private final long actualEnd;
public ISOCharSequenceBitVector(final CharSequence s, final boolean prefixFree) {
this.s = s;
actualEnd = s.length() * (long)Byte.SIZE;
length = actualEnd + (prefixFree ? Byte.SIZE : 0);
}
@Override
public boolean getBoolean(final long index) {
if (index > length) throw new IndexOutOfBoundsException();
if (index >= actualEnd) return false;
final int byteIndex = (int)(index >>> LOG2_BYTE_SIZE);
return (s.charAt(byteIndex) & 0x80 >>> (index & BYTE_MASK)) != 0;
}
@Override
public long getLong(final long from, final long to) {
final int startBit = (int)(from & BYTE_MASK);
if (startBit == 0) {
final int pos = (int)(from >>> LOG2_BYTE_SIZE);
if (to == from + Long.SIZE)
return reverseBytes((to > actualEnd ? 0 : (s.charAt(pos + 7) & 0xFFL)) << 56 |
(s.charAt(pos + 6) & 0xFFL) << 48 |
(s.charAt(pos + 5) & 0xFFL) << 40 |
(s.charAt(pos + 4) & 0xFFL) << 32 |
(s.charAt(pos + 3) & 0xFFL) << 24 |
(s.charAt(pos + 2) & 0xFF) << 16 |
(s.charAt(pos + 1) & 0xFF) << 8 |
(s.charAt(pos) & 0xFF));
if ((to & BYTE_MASK) == 0) {
long word = 0;
switch ((int)((Math.min(to, actualEnd) - Math.min(from, actualEnd)) >>> LOG2_BYTE_SIZE)) {
case 7: word |= (s.charAt(pos + 6) & 0xFFL) << 48;
case 6: word |= (s.charAt(pos + 5) & 0xFFL) << 40;
case 5: word |= (s.charAt(pos + 4) & 0xFFL) << 32;
case 4: word |= (s.charAt(pos + 3) & 0xFFL) << 24;
case 3: word |= (s.charAt(pos + 2) & 0xFF) << 16;
case 2: word |= (s.charAt(pos + 1) & 0xFF) << 8;
case 1: word |= s.charAt(pos) & 0xFF;
}
return reverseBytes(word);
}
}
final long l = Long.SIZE - (to - from);
final long startPos = from - startBit;
if (l == Long.SIZE) return 0;
if (startBit <= l) return getLong(startPos, Math.min(length, startPos + Long.SIZE)) << l - startBit >>> l;
return getLong(startPos, startPos + Long.SIZE) >>> startBit | getLong(startPos + Long.SIZE, Math.min(length, startPos + 2 * Long.SIZE)) << Long.SIZE + l - startBit >>> l;
}
@Override
public long length() {
return length;
}
}
private static class ISOMutableStringBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 1L;
private static final int LOG2_BYTE_SIZE = 3;
private static final long BYTE_MASK = Byte.SIZE - 1;
private final char[] a;
private final long length;
private final long actualEnd;
public ISOMutableStringBitVector(final MutableString s, final boolean prefixFree) {
this.a = s.array();
actualEnd = s.length() * (long)Byte.SIZE;
length = actualEnd + (prefixFree ? Byte.SIZE : 0);
}
@Override
public boolean getBoolean(final long index) {
if (index > length) throw new IndexOutOfBoundsException();
if (index >= actualEnd) return false;
final int byteIndex = (int)(index >>> LOG2_BYTE_SIZE);
return (a[byteIndex] & 0x80 >>> (index & BYTE_MASK)) != 0;
}
@Override
public long getLong(final long from, final long to) {
final int startBit = (int)(from & BYTE_MASK);
if (startBit == 0) {
final int pos = (int)(from >>> LOG2_BYTE_SIZE);
if (to == from + Long.SIZE)
return reverseBytes((to > actualEnd ? 0 : (a[pos + 7] & 0xFFL)) << 56 |
(a[pos + 6] & 0xFFL) << 48 |
(a[pos + 5] & 0xFFL) << 40 |
(a[pos + 4] & 0xFFL) << 32 |
(a[pos + 3] & 0xFFL) << 24 |
(a[pos + 2] & 0xFF) << 16 |
(a[pos + 1] & 0xFF) << 8 |
(a[pos] & 0xFF));
if ((to & BYTE_MASK) == 0) {
long word = 0;
switch ((int)((Math.min(actualEnd, to) - Math.min(actualEnd, from)) >>> LOG2_BYTE_SIZE)) {
case 7: word |= (a[pos + 6] & 0xFFL) << 48;
case 6: word |= (a[pos + 5] & 0xFFL) << 40;
case 5: word |= (a[pos + 4] & 0xFFL) << 32;
case 4: word |= (a[pos + 3] & 0xFFL) << 24;
case 3: word |= (a[pos + 2] & 0xFF) << 16;
case 2: word |= (a[pos + 1] & 0xFF) << 8;
case 1: word |= a[pos] & 0xFF;
}
return reverseBytes(word);
}
}
final long l = Long.SIZE - (to - from);
final long startPos = from - startBit;
if (l == Long.SIZE) return 0;
if (startBit <= l) return getLong(startPos, Math.min(length, startPos + Long.SIZE)) << l - startBit >>> l;
return getLong(startPos, startPos + Long.SIZE) >>> startBit | getLong(startPos + Long.SIZE, Math.min(length, startPos + 2 * Long.SIZE)) << Long.SIZE + l - startBit >>> l;
}
@Override
public long length() {
return length;
}
}
@Override
public BitVector toBitVector(final CharSequence s) {
return s instanceof MutableString ? new ISOMutableStringBitVector((MutableString)s, prefixFree) : new ISOCharSequenceBitVector(s, prefixFree);
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
private Object readResolve() {
return prefixFree ? PREFIX_FREE_ISO : ISO;
}
}
private static final TransformationStrategy RAW_BYTE_ARRAY = new RawByteArrayTransformationStrategy();
/** A trivial, high-performance, raw transformation from byte arrays to bit
* vectors that simply concatenates the bytes of the array.
*
* Warning: this transformation is not lexicographic.
*
*
Warning: bit vectors returned by this strategy are adaptors around the original array. If the array
* changes while the bit vector is being accessed, the results will be unpredictable.
*
* @see TransformationStrategies
*/
public static TransformationStrategy rawByteArray() {
return RAW_BYTE_ARRAY;
}
private static class RawByteArrayTransformationStrategy implements TransformationStrategy, Serializable {
private static final long serialVersionUID = 1L;
@Override
public long length(final byte[] a) {
return a.length * (long)Byte.SIZE;
}
private static class RawByteArrayBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 0L;
private static final int LOG2_BYTE_SIZE = 3;
private static final long BYTE_MASK = Byte.SIZE - 1;
private final byte[] a;
private final long length;
public RawByteArrayBitVector(final byte[] a) {
this.a = a;
length = a.length * (long)Byte.SIZE;
}
@Override
public boolean getBoolean(final long index) {
if (index > length) throw new IndexOutOfBoundsException();
return (a[(int)(index >>> LOG2_BYTE_SIZE)] & 1 << (index & BYTE_MASK)) != 0;
}
@Override
public long getLong(final long from, final long to) {
final int startBit = (int)(from & BYTE_MASK);
if (startBit == 0) {
if (to == from + Long.SIZE) {
final int pos = (int)(from >>> LOG2_BYTE_SIZE);
return (a[pos + 7] & 0xFFL) << 56 |
(a[pos + 6] & 0xFFL) << 48 |
(a[pos + 5] & 0xFFL) << 40 |
(a[pos + 4] & 0xFFL) << 32 |
(a[pos + 3] & 0xFFL) << 24 |
(a[pos + 2] & 0xFF) << 16 |
(a[pos + 1] & 0xFF) << 8 |
(a[pos] & 0xFF);
}
if ((to & BYTE_MASK) == 0) {
final int pos = (int)(from >>> LOG2_BYTE_SIZE);
long word = 0;
switch ((int)((to - from) >>> LOG2_BYTE_SIZE)) {
case 7: word |= (a[pos + 6] & 0xFFL) << 48;
case 6: word |= (a[pos + 5] & 0xFFL) << 40;
case 5: word |= (a[pos + 4] & 0xFFL) << 32;
case 4: word |= (a[pos + 3] & 0xFFL) << 24;
case 3: word |= (a[pos + 2] & 0xFF) << 16;
case 2: word |= (a[pos + 1] & 0xFF) << 8;
case 1: word |= (a[pos] & 0xFF);
}
return word;
}
}
// Actually, we should never get here as the transformation is not lexicographical.
final long l = Long.SIZE - (to - from);
final long startPos = from - startBit;
if (l == Long.SIZE) return 0;
if (startBit <= l) return getLong(startPos, Math.min(length, startPos + Long.SIZE)) << l - startBit >>> l;
return getLong(startPos, startPos + Long.SIZE) >>> startBit | getLong(startPos + Long.SIZE, Math.min(length, startPos + 2 * Long.SIZE)) << Long.SIZE + l - startBit >>> l;
}
@Override
public long length() {
return length;
}
}
@Override
public BitVector toBitVector(final byte[] s) {
return new RawByteArrayBitVector(s);
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
private Object readResolve() {
return RAW_BYTE_ARRAY;
}
}
private static final TransformationStrategy BYTE_ARRAY = new ByteArrayTransformationStrategy();
/** A lexicographical transformation from byte arrays to bit vectors.
*
* Warning: bit vectors returned by this strategy are adaptors around the original array. If the array
* changes while the bit vector is being accessed, the results will be unpredictable.
*
* @see TransformationStrategies
*/
public static TransformationStrategy byteArray() {
return BYTE_ARRAY;
}
private static class ByteArrayTransformationStrategy implements TransformationStrategy, Serializable {
private static final long serialVersionUID = 1L;
@Override
public long length(final byte[] a) {
return a.length * (long)Byte.SIZE;
}
private static class ByteArrayBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 0L;
private static final int LOG2_BYTE_SIZE = 3;
private static final long BYTE_MASK = Byte.SIZE - 1;
private final byte[] a;
private final long length;
public ByteArrayBitVector(final byte[] a) {
this.a = a;
length = a.length * (long)Byte.SIZE;
}
@Override
public boolean getBoolean(final long index) {
if (index > length) throw new IndexOutOfBoundsException();
return (a[(int)(index >>> LOG2_BYTE_SIZE)] & 0x80 >> (index & BYTE_MASK)) != 0;
}
@Override
public long getLong(final long from, final long to) {
final int startBit = (int)(from & BYTE_MASK);
if (startBit == 0) {
if (to == from + Long.SIZE) {
final int pos = (int)(from >>> LOG2_BYTE_SIZE);
return reverseBytes((a[pos + 7] & 0xFFL) << 56 |
(a[pos + 6] & 0xFFL) << 48 |
(a[pos + 5] & 0xFFL) << 40 |
(a[pos + 4] & 0xFFL) << 32 |
(a[pos + 3] & 0xFFL) << 24 |
(a[pos + 2] & 0xFF) << 16 |
(a[pos + 1] & 0xFF) << 8 |
(a[pos] & 0xFF));
}
if ((to & BYTE_MASK) == 0) {
final int pos = (int)(from >>> LOG2_BYTE_SIZE);
long word = 0;
switch ((int)((to - from) >>> LOG2_BYTE_SIZE)) {
case 7: word |= (a[pos + 6] & 0xFFL) << 48;
case 6: word |= (a[pos + 5] & 0xFFL) << 40;
case 5: word |= (a[pos + 4] & 0xFFL) << 32;
case 4: word |= (a[pos + 3] & 0xFFL) << 24;
case 3: word |= (a[pos + 2] & 0xFF) << 16;
case 2: word |= (a[pos + 1] & 0xFF) << 8;
case 1: word |= (a[pos] & 0xFF);
}
return reverseBytes(word);
}
}
// Actually, we should never get here as the transformation is not lexicographical.
final long l = Long.SIZE - (to - from);
final long startPos = from - startBit;
if (l == Long.SIZE) return 0;
if (startBit <= l) return getLong(startPos, Math.min(length, startPos + Long.SIZE)) << l - startBit >>> l;
return getLong(startPos, startPos + Long.SIZE) >>> startBit | getLong(startPos + Long.SIZE, Math.min(length, startPos + 2 * Long.SIZE)) << Long.SIZE + l - startBit >>> l;
}
@Override
public long length() {
return length;
}
}
@Override
public BitVector toBitVector(final byte[] s) {
return new ByteArrayBitVector(s);
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
private Object readResolve() {
return BYTE_ARRAY;
}
}
private final static class IteratorWrapper implements ObjectIterator {
final Iterator iterator;
final TransformationStrategy transformationStrategy;
public IteratorWrapper(final Iterator iterator, final TransformationStrategy transformationStrategy) {
this.iterator = iterator;
this.transformationStrategy = transformationStrategy;
}
@Override
public boolean hasNext() {
return iterator.hasNext();
}
@Override
public BitVector next() {
return transformationStrategy.toBitVector(iterator.next());
}
}
/** Wraps a given iterator, returning an iterator that emits {@linkplain BitVector bit vectors}.
*
* @param iterator an iterator.
* @param transformationStrategy a strategy to transform the object returned by iterator.
* @return an iterator that emits the content of iterator passed through transformationStrategy.
*/
@SuppressWarnings("unchecked")
public static Iterator wrap(final Iterator iterator, final TransformationStrategy transformationStrategy) {
return (Iterator)(transformationStrategy == IDENTITY ? iterator : new IteratorWrapper<>(iterator, transformationStrategy));
}
private final static class IterableWrapper implements Iterable {
private final TransformationStrategy transformationStrategy;
private final Iterable collection;
public IterableWrapper(final Iterable collection, final TransformationStrategy transformationStrategy) {
this.collection = collection;
this.transformationStrategy = transformationStrategy;
}
@Override
public ObjectIterator iterator() {
return new IteratorWrapper<>(collection.iterator(), transformationStrategy.copy());
}
}
/** Wraps a given iterable, returning an iterable that contains {@linkplain BitVector bit vectors}.
*
* @param iterable an iterable.
* @param transformationStrategy a strategy to transform the object contained in iterable.
* @return an iterable that has the content of iterable passed through transformationStrategy.
*/
@SuppressWarnings("unchecked")
public static Iterable wrap(final Iterable iterable, final TransformationStrategy transformationStrategy) {
return (Iterable)(transformationStrategy == IDENTITY ? iterable : new IterableWrapper<>(iterable, transformationStrategy));
}
private final static class ListWrapper extends AbstractObjectList {
private final TransformationStrategy transformationStrategy;
private final List list;
public ListWrapper(final List list, final TransformationStrategy transformationStrategy) {
this.list = list;
this.transformationStrategy = transformationStrategy;
}
@Override
public BitVector get(final int index) {
return transformationStrategy.toBitVector(list.get(index));
}
@Override
public int size() {
return list.size();
}
}
/** Wraps a given list, returning a list that contains {@linkplain BitVector bit vectors}.
*
* @param list a list.
* @param transformationStrategy a strategy to transform the object contained in list.
* @return a list that has the content of list passed through transformationStrategy.
*/
@SuppressWarnings("unchecked")
public static List wrap(final List list, final TransformationStrategy transformationStrategy) {
return (List)(transformationStrategy == IDENTITY ? list : new ListWrapper<>(list, transformationStrategy));
}
private static final TransformationStrategy PREFIX_FREE = new PrefixFreeTransformationStrategy();
/** A transformation from bit vectors to bit vectors that guarantees that its results are prefix free.
*
* More in detail, we map 0 to 10, 1 to 11, and we add a 0 at the end of all strings.
*
*
Warning: bit vectors returned by this strategy are adaptors around the original string. If the string
* changes while the bit vector is being accessed, the results will be unpredictable.
*/
@SuppressWarnings("unchecked")
public static TransformationStrategy prefixFree() {
return (TransformationStrategy)PREFIX_FREE;
}
private static class PrefixFreeTransformationStrategy implements TransformationStrategy, Serializable {
private static final long serialVersionUID = 1L;
private static class PrefixFreeBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 1L;
private final BitVector v;
private final long length;
public PrefixFreeBitVector(final BitVector v) {
this.v = v;
length = (v.length() << 1) + 1;
}
@Override
public boolean getBoolean(final long index) {
if (index >= length) throw new IndexOutOfBoundsException();
if (index == length - 1) return false;
if ((index & 1) == 0) return true;
return v.getBoolean(index >> 1);
}
@Override
public long getLong(final long from, final long to) {
// The following code is optimized for word-by-word reading.
if (LongArrayBitVector.round(from)) {
if (to == from + Long.SIZE) {
long word = v.getLong(from >> 1, (from >> 1) + Long.SIZE / 2);
word = (word | word << 16) & 0x0000FFFF0000FFFFL;
word = (word | word << 8) & 0x00FF00FF00FF00FFL;
word = (word | word << 4) & 0x0F0F0F0F0F0F0F0FL;
word = (word | word << 2) & 0x3333333333333333L;
return word << 1 | word << 2 | 0x5555555555555555L;
}
if (to == length) {
assert from < to; // As from is even and to is odd.
long word = v.getLong(from >> 1, to >> 1);
word = (word | word << 16) & 0x0000FFFF0000FFFFL;
word = (word | word << 8) & 0x00FF00FF00FF00FFL;
word = (word | word << 4) & 0x0F0F0F0F0F0F0F0FL;
word = (word | word << 2) & 0x3333333333333333L;
return (word << 1 | word << 2 | 0x5555555555555555L) & (1L << to - from - 1) - 1;
}
}
return super.getLong(from, to);
}
@Override
public long length() {
return length;
}
}
@Override
public BitVector toBitVector(final BitVector v) {
return new PrefixFreeBitVector(v);
}
@Override
public long length(final BitVector v) {
return v.length() * 2 + 1;
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
private Object readResolve() {
return PREFIX_FREE;
}
}
private static final FixedLongTransformationStrategy FIXED_LONG = new FixedLongTransformationStrategy(true);
/** A transformation from longs to bit vectors that returns a fixed-size {@link Long#SIZE}-bit vector. Note that the
* bit vectors have as first bit the most significant bit of the underlying long integer, so
* lexicographical and numerical order do coincide for positive numbers. */
public static TransformationStrategy fixedLong() {
return FIXED_LONG;
}
private static final FixedLongTransformationStrategy RAW_FIXED_LONG = new FixedLongTransformationStrategy(false);
/** A trivial, high-performance, raw transformation from longs to bit vectors that returns a fixed-size
* {@link Long#SIZE}-bit vector. */
public static TransformationStrategy rawFixedLong() {
return RAW_FIXED_LONG;
}
/** A transformation from longs to bit vectors that returns a fixed-size {@link Long#SIZE}-bit vector, possibly reversed
* to maintain lexicographical order. */
private static class FixedLongTransformationStrategy implements TransformationStrategy, Serializable {
private static final long serialVersionUID = 0L;
private final boolean lexicographical;
public FixedLongTransformationStrategy(final boolean lexicographical) {
this.lexicographical = lexicographical;
}
private static class FixedLongBitVector extends AbstractBitVector implements Serializable {
private static final long serialVersionUID = 1L;
private final long v;
public FixedLongBitVector(final long v) {
this.v = v;
}
@Override
public boolean getBoolean(final long index) {
if (index >= Long.SIZE) throw new IndexOutOfBoundsException();
return (v & 1L << index) != 0;
}
@Override
public long getLong(final long from, final long to) {
if (from == 0 && to == Long.SIZE) return v;
return (v >> from) & (1L << to - from) - 1;
}
@Override
public long length() {
return Long.SIZE;
}
}
@Override
public BitVector toBitVector(final Long v) {
return new FixedLongBitVector(lexicographical ? Long.reverse(v.longValue()) : v.longValue());
}
@Override
public long length(final Long v) {
return Long.SIZE;
}
@Override
public long numBits() { return 0; }
@Override
public TransformationStrategy copy() {
return this;
}
private Object readResolve() {
return FIXED_LONG;
}
}
}