it.unimi.dsi.fastutil.doubles.DoubleOpenHashBigSet Maven / Gradle / Ivy
Show all versions of fastutil Show documentation
/*
* Copyright (C) 2002-2017 Sebastiano Vigna
*
* Licensed 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.
*/
package it.unimi.dsi.fastutil.doubles;
import it.unimi.dsi.fastutil.BigArrays;
import it.unimi.dsi.fastutil.Hash;
import it.unimi.dsi.fastutil.Size64;
import it.unimi.dsi.fastutil.HashCommon;
import static it.unimi.dsi.fastutil.HashCommon.bigArraySize;
import static it.unimi.dsi.fastutil.HashCommon.maxFill;
import java.util.Collection;
import java.util.Iterator;
import java.util.NoSuchElementException;
/**
* A type-specific hash big set with with a fast, small-footprint
* implementation.
*
*
* Instances of this class use a hash table to represent a big set: the number
* of elements in the set is limited only by the amount of core memory. The
* table (backed by a {@linkplain it.unimi.dsi.fastutil.BigArrays big array}) is
* filled up to a specified load factor, and then doubled in size to
* accommodate new entries. If the table is emptied below one fourth of
* the load factor, it is halved in size; however, the table is never reduced to
* a size smaller than that at creation time: this approach makes it possible to
* create sets with a large capacity in which insertions and deletions do not
* cause immediately rehashing. Moreover, halving is not performed when deleting
* entries from an iterator, as it would interfere with the iteration process.
*
*
* Note that {@link #clear()} does not modify the hash table size. Rather, a
* family of {@linkplain #trim() trimming methods} lets you control the size of
* the table; this is particularly useful if you reuse instances of this class.
*
*
* The methods of this class are about 30% slower than those of the
* corresponding non-big set.
*
* @see Hash
* @see HashCommon
*/
public class DoubleOpenHashBigSet extends AbstractDoubleSet implements java.io.Serializable, Cloneable, Hash, Size64 {
private static final long serialVersionUID = 0L;
private static final boolean ASSERTS = false;
/** The big array of keys. */
protected transient double[][] key;
/** The mask for wrapping a position counter. */
protected transient long mask;
/** The mask for wrapping a segment counter. */
protected transient int segmentMask;
/** The mask for wrapping a base counter. */
protected transient int baseMask;
/** Whether this set contains the null key. */
protected transient boolean containsNull;
/** The current table size (always a power of 2). */
protected transient long n;
/**
* Threshold after which we rehash. It must be the table size times {@link #f}.
*/
protected transient long maxFill;
/**
* We never resize below this threshold, which is the construction-time {#n}.
*/
protected final transient long minN;
/** The acceptable load factor. */
protected final float f;
/** Number of entries in the set. */
protected long size;
/** Initialises the mask values. */
private void initMasks() {
mask = n - 1;
/*
* Note that either we have more than one segment, and in this case all segments
* are BigArrays.SEGMENT_SIZE long, or we have exactly one segment whose length
* is a power of two.
*/
segmentMask = key[0].length - 1;
baseMask = key.length - 1;
}
/**
* Creates a new hash big set.
*
*
* The actual table size will be the least power of two greater than
* {@code expected}/{@code f}.
*
* @param expected
* the expected number of elements in the set.
* @param f
* the load factor.
*/
public DoubleOpenHashBigSet(final long expected, final float f) {
if (f <= 0 || f > 1)
throw new IllegalArgumentException("Load factor must be greater than 0 and smaller than or equal to 1");
if (n < 0)
throw new IllegalArgumentException("The expected number of elements must be nonnegative");
this.f = f;
minN = n = bigArraySize(expected, f);
maxFill = maxFill(n, f);
key = DoubleBigArrays.newBigArray(n);
initMasks();
}
/**
* Creates a new hash big set with {@link Hash#DEFAULT_LOAD_FACTOR} as load
* factor.
*
* @param expected
* the expected number of elements in the hash big set.
*/
public DoubleOpenHashBigSet(final long expected) {
this(expected, DEFAULT_LOAD_FACTOR);
}
/**
* Creates a new hash big set with initial expected
* {@link Hash#DEFAULT_INITIAL_SIZE} elements and
* {@link Hash#DEFAULT_LOAD_FACTOR} as load factor.
*/
public DoubleOpenHashBigSet() {
this(DEFAULT_INITIAL_SIZE, DEFAULT_LOAD_FACTOR);
}
/**
* Creates a new hash big set copying a given collection.
*
* @param c
* a {@link Collection} to be copied into the new hash big set.
* @param f
* the load factor.
*/
public DoubleOpenHashBigSet(final Collection c, final float f) {
this(c.size(), f);
addAll(c);
}
/**
* Creates a new hash big set with {@link Hash#DEFAULT_LOAD_FACTOR} as load
* factor copying a given collection.
*
* @param c
* a {@link Collection} to be copied into the new hash big set.
*/
public DoubleOpenHashBigSet(final Collection c) {
this(c, DEFAULT_LOAD_FACTOR);
}
/**
* Creates a new hash big set copying a given type-specific collection.
*
* @param c
* a type-specific collection to be copied into the new hash big set.
* @param f
* the load factor.
*/
public DoubleOpenHashBigSet(final DoubleCollection c, final float f) {
this(c.size(), f);
addAll(c);
}
/**
* Creates a new hash big set with {@link Hash#DEFAULT_LOAD_FACTOR} as load
* factor copying a given type-specific collection.
*
* @param c
* a type-specific collection to be copied into the new hash big set.
*/
public DoubleOpenHashBigSet(final DoubleCollection c) {
this(c, DEFAULT_LOAD_FACTOR);
}
/**
* Creates a new hash big set using elements provided by a type-specific
* iterator.
*
* @param i
* a type-specific iterator whose elements will fill the new hash big
* set.
* @param f
* the load factor.
*/
public DoubleOpenHashBigSet(final DoubleIterator i, final float f) {
this(DEFAULT_INITIAL_SIZE, f);
while (i.hasNext())
add(i.nextDouble());
}
/**
* Creates a new hash big set with {@link Hash#DEFAULT_LOAD_FACTOR} as load
* factor using elements provided by a type-specific iterator.
*
* @param i
* a type-specific iterator whose elements will fill the new hash big
* set.
*/
public DoubleOpenHashBigSet(final DoubleIterator i) {
this(i, DEFAULT_LOAD_FACTOR);
}
/**
* Creates a new hash big set using elements provided by an iterator.
*
* @param i
* an iterator whose elements will fill the new hash big set.
* @param f
* the load factor.
*/
public DoubleOpenHashBigSet(final Iterator i, final float f) {
this(DoubleIterators.asDoubleIterator(i), f);
}
/**
* Creates a new hash big set with {@link Hash#DEFAULT_LOAD_FACTOR} as load
* factor using elements provided by an iterator.
*
* @param i
* an iterator whose elements will fill the new hash big set.
*/
public DoubleOpenHashBigSet(final Iterator i) {
this(DoubleIterators.asDoubleIterator(i));
}
/**
* Creates a new hash big set and fills it with the elements of a given array.
*
* @param a
* an array whose elements will be used to fill the new hash big set.
* @param offset
* the first element to use.
* @param length
* the number of elements to use.
* @param f
* the load factor.
*/
public DoubleOpenHashBigSet(final double[] a, final int offset, final int length, final float f) {
this(length < 0 ? 0 : length, f);
DoubleArrays.ensureOffsetLength(a, offset, length);
for (int i = 0; i < length; i++)
add(a[offset + i]);
}
/**
* Creates a new hash big set with {@link Hash#DEFAULT_LOAD_FACTOR} as load
* factor and fills it with the elements of a given array.
*
* @param a
* an array whose elements will be used to fill the new hash big set.
* @param offset
* the first element to use.
* @param length
* the number of elements to use.
*/
public DoubleOpenHashBigSet(final double[] a, final int offset, final int length) {
this(a, offset, length, DEFAULT_LOAD_FACTOR);
}
/**
* Creates a new hash big set copying the elements of an array.
*
* @param a
* an array to be copied into the new hash big set.
* @param f
* the load factor.
*/
public DoubleOpenHashBigSet(final double[] a, final float f) {
this(a, 0, a.length, f);
}
/**
* Creates a new hash big set with {@link Hash#DEFAULT_LOAD_FACTOR} as load
* factor copying the elements of an array.
*
* @param a
* an array to be copied into the new hash big set.
*/
public DoubleOpenHashBigSet(final double[] a) {
this(a, DEFAULT_LOAD_FACTOR);
}
private long realSize() {
return containsNull ? size - 1 : size;
}
private void ensureCapacity(final long capacity) {
final long needed = bigArraySize(capacity, f);
if (needed > n)
rehash(needed);
}
@Override
public boolean addAll(Collection c) {
final long size = c instanceof Size64 ? ((Size64) c).size64() : c.size();
// The resulting collection will be at least c.size() big
if (f <= .5)
ensureCapacity(size); // The resulting collection will be sized for c.size() elements
else
ensureCapacity(size64() + size); // The resulting collection will be sized for size() + c.size() elements
return super.addAll(c);
}
@Override
public boolean addAll(DoubleCollection c) {
final long size = c instanceof Size64 ? ((Size64) c).size64() : c.size();
if (f <= .5)
ensureCapacity(size); // The resulting collection will be size for c.size() elements
else
ensureCapacity(size64() + size); // The resulting collection will be sized for size() + c.size() elements
return super.addAll(c);
}
@Override
public boolean add(final double k) {
int displ, base;
if ((Double.doubleToLongBits(k) == 0)) {
if (containsNull)
return false;
containsNull = true;
} else {
double curr;
final double[][] key = this.key;
final long h = it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k));
// The starting point.
if (!(Double
.doubleToLongBits(curr = key[base = (int) ((h & mask) >>> BigArrays.SEGMENT_SHIFT)][displ = (int) (h
& segmentMask)]) == 0)) {
if ((Double.doubleToLongBits(curr) == Double.doubleToLongBits(k)))
return false;
while (!(Double
.doubleToLongBits(curr = key[base = (base + ((displ = (displ + 1) & segmentMask) == 0 ? 1 : 0))
& baseMask][displ]) == 0))
if ((Double.doubleToLongBits(curr) == Double.doubleToLongBits(k)))
return false;
}
key[base][displ] = k;
}
if (size++ >= maxFill)
rehash(2 * n);
if (ASSERTS)
checkTable();
return true;
}
/**
* Shifts left entries with the specified hash code, starting at the specified
* position, and empties the resulting free entry.
*
* @param pos
* a starting position.
*/
protected final void shiftKeys(long pos) {
// Shift entries with the same hash.
long last, slot;
final double[][] key = this.key;
for (;;) {
pos = ((last = pos) + 1) & mask;
for (;;) {
if ((Double.doubleToLongBits(DoubleBigArrays.get(key, pos)) == 0)) {
DoubleBigArrays.set(key, last, (0));
return;
}
slot = it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(DoubleBigArrays.get(key, pos)))
& mask;
if (last <= pos ? last >= slot || slot > pos : last >= slot && slot > pos)
break;
pos = (pos + 1) & mask;
}
DoubleBigArrays.set(key, last, DoubleBigArrays.get(key, pos));
}
}
private boolean removeEntry(final int base, final int displ) {
size--;
shiftKeys(base * (long) BigArrays.SEGMENT_SIZE + displ);
if (n > minN && size < maxFill / 4 && n > DEFAULT_INITIAL_SIZE)
rehash(n / 2);
return true;
}
private boolean removeNullEntry() {
containsNull = false;
size--;
if (n > minN && size < maxFill / 4 && n > DEFAULT_INITIAL_SIZE)
rehash(n / 2);
return true;
}
@Override
public boolean remove(final double k) {
if ((Double.doubleToLongBits(k) == 0)) {
if (containsNull)
return removeNullEntry();
return false;
}
double curr;
final double[][] key = this.key;
final long h = it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k));
int displ, base;
// The starting point.
if ((Double.doubleToLongBits(curr = key[base = (int) ((h & mask) >>> BigArrays.SEGMENT_SHIFT)][displ = (int) (h
& segmentMask)]) == 0))
return false;
if ((Double.doubleToLongBits(curr) == Double.doubleToLongBits(k)))
return removeEntry(base, displ);
while (true) {
if ((Double.doubleToLongBits(curr = key[base = (base + ((displ = (displ + 1) & segmentMask) == 0 ? 1 : 0))
& baseMask][displ]) == 0))
return false;
if ((Double.doubleToLongBits(curr) == Double.doubleToLongBits(k)))
return removeEntry(base, displ);
}
}
@Override
public boolean contains(final double k) {
if ((Double.doubleToLongBits(k) == 0))
return containsNull;
double curr;
final double[][] key = this.key;
final long h = it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k));
int displ, base;
// The starting point.
if ((Double.doubleToLongBits(curr = key[base = (int) ((h & mask) >>> BigArrays.SEGMENT_SHIFT)][displ = (int) (h
& segmentMask)]) == 0))
return false;
if ((Double.doubleToLongBits(curr) == Double.doubleToLongBits(k)))
return true;
while (true) {
if ((Double.doubleToLongBits(curr = key[base = (base + ((displ = (displ + 1) & segmentMask) == 0 ? 1 : 0))
& baseMask][displ]) == 0))
return false;
if ((Double.doubleToLongBits(curr) == Double.doubleToLongBits(k)))
return true;
}
}
/*
* Removes all elements from this set.
*
*/
/**
* {@inheritDoc}
*
*
* To increase object reuse, this method does not change the table size. If you
* want to reduce the table size, you must use {@link #trim(long)}.
*/
@Override
public void clear() {
if (size == 0)
return;
size = 0;
containsNull = false;
DoubleBigArrays.fill(key, (0));
}
/** An iterator over a hash big set. */
private class SetIterator implements DoubleIterator {
/**
* The base of the last entry returned, if positive or zero; initially, the
* number of components of the key array. If negative, the last element returned
* was that of index {@code - base - 1} from the {@link #wrapped} list.
*/
int base = key.length;
/** The displacement of the last entry returned; initially, zero. */
int displ;
/**
* The index of the last entry that has been returned (or {@link Long#MIN_VALUE}
* if {@link #base} is negative). It is -1 if either we did not return an entry
* yet, or the last returned entry has been removed.
*/
long last = -1;
/** A downward counter measuring how many entries must still be returned. */
long c = size;
/** A boolean telling us whether we should return the null key. */
boolean mustReturnNull = DoubleOpenHashBigSet.this.containsNull;
/**
* A lazily allocated list containing elements that have wrapped around the
* table because of removals.
*/
DoubleArrayList wrapped;
@Override
public boolean hasNext() {
return c != 0;
}
@Override
public double nextDouble() {
if (!hasNext())
throw new NoSuchElementException();
c--;
if (mustReturnNull) {
mustReturnNull = false;
last = n;
return (0);
}
final double[][] key = DoubleOpenHashBigSet.this.key;
for (;;) {
if (displ == 0 && base <= 0) {
// We are just enumerating elements from the wrapped list.
last = Long.MIN_VALUE;
return wrapped.getDouble(-(--base) - 1);
}
if (displ-- == 0)
displ = key[--base].length - 1;
final double k = key[base][displ];
if (!(Double.doubleToLongBits(k) == 0)) {
last = base * (long) BigArrays.SEGMENT_SIZE + displ;
return k;
}
}
}
/**
* Shifts left entries with the specified hash code, starting at the specified
* position, and empties the resulting free entry.
*
* @param pos
* a starting position.
*/
private final void shiftKeys(long pos) {
// Shift entries with the same hash.
long last, slot;
double curr;
final double[][] key = DoubleOpenHashBigSet.this.key;
for (;;) {
pos = ((last = pos) + 1) & mask;
for (;;) {
if ((Double.doubleToLongBits(curr = DoubleBigArrays.get(key, pos)) == 0)) {
DoubleBigArrays.set(key, last, (0));
return;
}
slot = it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(curr)) & mask;
if (last <= pos ? last >= slot || slot > pos : last >= slot && slot > pos)
break;
pos = (pos + 1) & mask;
}
if (pos < last) { // Wrapped entry.
if (wrapped == null)
wrapped = new DoubleArrayList();
wrapped.add(DoubleBigArrays.get(key, pos));
}
DoubleBigArrays.set(key, last, curr);
}
}
@Override
public void remove() {
if (last == -1)
throw new IllegalStateException();
if (last == n)
DoubleOpenHashBigSet.this.containsNull = false;
else if (base >= 0)
shiftKeys(last);
else {
// We're removing wrapped entries.
DoubleOpenHashBigSet.this.remove(wrapped.getDouble(-base - 1));
last = -1; // Note that we must not decrement size
return;
}
size--;
last = -1; // You can no longer remove this entry.
if (ASSERTS)
checkTable();
}
}
@Override
public DoubleIterator iterator() {
return new SetIterator();
}
/**
* Rehashes this set, making the table as small as possible.
*
*
* This method rehashes the table to the smallest size satisfying the load
* factor. It can be used when the set will not be changed anymore, so to
* optimize access speed and size.
*
*
* If the table size is already the minimum possible, this method does nothing.
*
* @return true if there was enough memory to trim the set.
* @see #trim(long)
*/
public boolean trim() {
final long l = bigArraySize(size, f);
if (l >= n || size > maxFill(l, f))
return true;
try {
rehash(l);
} catch (OutOfMemoryError cantDoIt) {
return false;
}
return true;
}
/**
* Rehashes this set if the table is too large.
*
*
* Let N be the smallest table size that can hold
* max(n,{@link #size64()}) entries, still satisfying the load
* factor. If the current table size is smaller than or equal to N,
* this method does nothing. Otherwise, it rehashes this set in a table of size
* N.
*
*
* This method is useful when reusing sets. {@linkplain #clear() Clearing a set}
* leaves the table size untouched. If you are reusing a set many times, you can
* call this method with a typical size to avoid keeping around a very large
* table just because of a few large transient sets.
*
* @param n
* the threshold for the trimming.
* @return true if there was enough memory to trim the set.
* @see #trim()
*/
public boolean trim(final long n) {
final long l = bigArraySize(n, f);
if (this.n <= l)
return true;
try {
rehash(l);
} catch (OutOfMemoryError cantDoIt) {
return false;
}
return true;
}
/**
* Resizes the set.
*
*
* This method implements the basic rehashing strategy, and may be overriden by
* subclasses implementing different rehashing strategies (e.g., disk-based
* rehashing). However, you should not override this method unless you
* understand the internal workings of this class.
*
* @param newN
* the new size
*/
protected void rehash(final long newN) {
final double key[][] = this.key;
final double newKey[][] = DoubleBigArrays.newBigArray(newN);
final long mask = newN - 1; // Note that this is used by the hashing macro
final int newSegmentMask = newKey[0].length - 1;
final int newBaseMask = newKey.length - 1;
int base = 0, displ = 0, b, d;
long h;
double k;
for (long i = realSize(); i-- != 0;) {
while ((Double.doubleToLongBits(key[base][displ]) == 0))
base = (base + ((displ = (displ + 1) & segmentMask) == 0 ? 1 : 0));
k = key[base][displ];
h = it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k));
// The starting point.
if (!(Double.doubleToLongBits(
newKey[b = (int) ((h & mask) >>> BigArrays.SEGMENT_SHIFT)][d = (int) (h & newSegmentMask)]) == 0))
while (!(Double.doubleToLongBits(
newKey[b = (b + ((d = (d + 1) & newSegmentMask) == 0 ? 1 : 0)) & newBaseMask][d]) == 0));
newKey[b][d] = k;
base = (base + ((displ = (displ + 1) & segmentMask) == 0 ? 1 : 0));
}
this.n = newN;
this.key = newKey;
initMasks();
maxFill = maxFill(n, f);
}
@Deprecated
@Override
public int size() {
return (int) Math.min(Integer.MAX_VALUE, size);
}
@Override
public long size64() {
return size;
}
@Override
public boolean isEmpty() {
return size == 0;
}
/**
* Returns a deep copy of this big set.
*
*
* This method performs a deep copy of this big hash set; the data stored in the
* set, however, is not cloned. Note that this makes a difference only for
* object keys.
*
* @return a deep copy of this big set.
*/
@Override
public DoubleOpenHashBigSet clone() {
DoubleOpenHashBigSet c;
try {
c = (DoubleOpenHashBigSet) super.clone();
} catch (CloneNotSupportedException cantHappen) {
throw new InternalError();
}
c.key = DoubleBigArrays.copy(key);
c.containsNull = containsNull;
return c;
}
/**
* Returns a hash code for this set.
*
* This method overrides the generic method provided by the superclass. Since
* {@code equals()} is not overriden, it is important that the value returned by
* this method is the same value as the one returned by the overriden method.
*
* @return a hash code for this set.
*/
@Override
public int hashCode() {
final double key[][] = this.key;
int h = 0, base = 0, displ = 0;
for (long j = realSize(); j-- != 0;) {
while ((Double.doubleToLongBits(key[base][displ]) == 0))
base = (base + ((displ = (displ + 1) & segmentMask) == 0 ? 1 : 0));
h += it.unimi.dsi.fastutil.HashCommon.double2int(key[base][displ]);
base = (base + ((displ = (displ + 1) & segmentMask) == 0 ? 1 : 0));
}
return h;
}
private void writeObject(java.io.ObjectOutputStream s) throws java.io.IOException {
final DoubleIterator i = iterator();
s.defaultWriteObject();
for (long j = size; j-- != 0;)
s.writeDouble(i.nextDouble());
}
private void readObject(java.io.ObjectInputStream s) throws java.io.IOException, ClassNotFoundException {
s.defaultReadObject();
n = bigArraySize(size, f);
maxFill = maxFill(n, f);
final double[][] key = this.key = DoubleBigArrays.newBigArray(n);
initMasks();
long h;
double k;
int base, displ;
for (long i = size; i-- != 0;) {
k = s.readDouble();
if ((Double.doubleToLongBits(k) == 0))
containsNull = true;
else {
h = it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k));
if (!(Double
.doubleToLongBits(key[base = (int) ((h & mask) >>> BigArrays.SEGMENT_SHIFT)][displ = (int) (h
& segmentMask)]) == 0))
while (!(Double
.doubleToLongBits(key[base = (base + ((displ = (displ + 1) & segmentMask) == 0 ? 1 : 0))
& baseMask][displ]) == 0));
key[base][displ] = k;
}
}
if (ASSERTS)
checkTable();
}
private void checkTable() {
}
}