All Downloads are FREE. Search and download functionalities are using the official Maven repository.

it.unimi.dsi.fastutil.doubles.Double2DoubleOpenHashMap Maven / Gradle / Ivy

Go to download

fastutil extends the Java Collections Framework by providing type-specific maps, sets, lists, and queues with a small memory footprint and fast operations; it provides also big (64-bit) arrays, sets, and lists, sorting algorithms, fast, practical I/O classes for binary and text files, and facilities for memory mapping large files. This jar (fastutil-core.jar) contains data structures based on integers, longs, doubles, and objects, only; fastutil.jar contains all classes. If you have both jars in your dependencies, this jar should be excluded.

There is a newer version: 8.5.15
Show newest version
/*
	* Copyright (C) 2002-2021 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.Hash;
import it.unimi.dsi.fastutil.HashCommon;
import static it.unimi.dsi.fastutil.HashCommon.arraySize;
import static it.unimi.dsi.fastutil.HashCommon.maxFill;
import java.util.Map;
import java.util.Arrays;
import java.util.NoSuchElementException;
import java.util.function.Consumer;
import it.unimi.dsi.fastutil.objects.AbstractObjectSet;
import it.unimi.dsi.fastutil.objects.ObjectIterator;
import it.unimi.dsi.fastutil.objects.ObjectSpliterator;
import it.unimi.dsi.fastutil.objects.ObjectSpliterators;
/**
 * A type-specific hash map with a fast, small-footprint implementation.
 *
 * 

* Instances of this class use a hash table to represent a map. The table 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 maps 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. * *

* Entries returned by the type-specific {@link #entrySet()} method implement * the suitable type-specific {@link it.unimi.dsi.fastutil.Pair Pair} interface; * only values are mutable. * * @see Hash * @see HashCommon */ public class Double2DoubleOpenHashMap extends AbstractDouble2DoubleMap implements java.io.Serializable, Cloneable, Hash { private static final long serialVersionUID = 0L; private static final boolean ASSERTS = false; /** The array of keys. */ protected transient double[] key; /** The array of values. */ protected transient double[] value; /** The mask for wrapping a position counter. */ protected transient int mask; /** Whether this map contains the key zero. */ protected transient boolean containsNullKey; /** The current table size. */ protected transient int n; /** * Threshold after which we rehash. It must be the table size times {@link #f}. */ protected transient int maxFill; /** * We never resize below this threshold, which is the construction-time {#n}. */ protected final transient int minN; /** Number of entries in the set (including the key zero, if present). */ protected int size; /** The acceptable load factor. */ protected final float f; /** Cached set of entries. */ protected transient FastEntrySet entries; /** Cached set of keys. */ protected transient DoubleSet keys; /** Cached collection of values. */ protected transient DoubleCollection values; /** * Creates a new hash map. * *

* 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 hash map. * @param f * the load factor. */ public Double2DoubleOpenHashMap(final int expected, final float f) { if (f <= 0 || f >= 1) throw new IllegalArgumentException("Load factor must be greater than 0 and smaller than 1"); if (expected < 0) throw new IllegalArgumentException("The expected number of elements must be nonnegative"); this.f = f; minN = n = arraySize(expected, f); mask = n - 1; maxFill = maxFill(n, f); key = new double[n + 1]; value = new double[n + 1]; } /** * Creates a new hash map with {@link Hash#DEFAULT_LOAD_FACTOR} as load factor. * * @param expected * the expected number of elements in the hash map. */ public Double2DoubleOpenHashMap(final int expected) { this(expected, DEFAULT_LOAD_FACTOR); } /** * Creates a new hash map with initial expected * {@link Hash#DEFAULT_INITIAL_SIZE} entries and * {@link Hash#DEFAULT_LOAD_FACTOR} as load factor. */ public Double2DoubleOpenHashMap() { this(DEFAULT_INITIAL_SIZE, DEFAULT_LOAD_FACTOR); } /** * Creates a new hash map copying a given one. * * @param m * a {@link Map} to be copied into the new hash map. * @param f * the load factor. */ public Double2DoubleOpenHashMap(final Map m, final float f) { this(m.size(), f); putAll(m); } /** * Creates a new hash map with {@link Hash#DEFAULT_LOAD_FACTOR} as load factor * copying a given one. * * @param m * a {@link Map} to be copied into the new hash map. */ public Double2DoubleOpenHashMap(final Map m) { this(m, DEFAULT_LOAD_FACTOR); } /** * Creates a new hash map copying a given type-specific one. * * @param m * a type-specific map to be copied into the new hash map. * @param f * the load factor. */ public Double2DoubleOpenHashMap(final Double2DoubleMap m, final float f) { this(m.size(), f); putAll(m); } /** * Creates a new hash map with {@link Hash#DEFAULT_LOAD_FACTOR} as load factor * copying a given type-specific one. * * @param m * a type-specific map to be copied into the new hash map. */ public Double2DoubleOpenHashMap(final Double2DoubleMap m) { this(m, DEFAULT_LOAD_FACTOR); } /** * Creates a new hash map using the elements of two parallel arrays. * * @param k * the array of keys of the new hash map. * @param v * the array of corresponding values in the new hash map. * @param f * the load factor. * @throws IllegalArgumentException * if {@code k} and {@code v} have different lengths. */ public Double2DoubleOpenHashMap(final double[] k, final double[] v, final float f) { this(k.length, f); if (k.length != v.length) throw new IllegalArgumentException( "The key array and the value array have different lengths (" + k.length + " and " + v.length + ")"); for (int i = 0; i < k.length; i++) this.put(k[i], v[i]); } /** * Creates a new hash map with {@link Hash#DEFAULT_LOAD_FACTOR} as load factor * using the elements of two parallel arrays. * * @param k * the array of keys of the new hash map. * @param v * the array of corresponding values in the new hash map. * @throws IllegalArgumentException * if {@code k} and {@code v} have different lengths. */ public Double2DoubleOpenHashMap(final double[] k, final double[] v) { this(k, v, DEFAULT_LOAD_FACTOR); } private int realSize() { return containsNullKey ? size - 1 : size; } private void ensureCapacity(final int capacity) { final int needed = arraySize(capacity, f); if (needed > n) rehash(needed); } private void tryCapacity(final long capacity) { final int needed = (int) Math.min(1 << 30, Math.max(2, HashCommon.nextPowerOfTwo((long) Math.ceil(capacity / f)))); if (needed > n) rehash(needed); } private double removeEntry(final int pos) { final double oldValue = value[pos]; size--; shiftKeys(pos); if (n > minN && size < maxFill / 4 && n > DEFAULT_INITIAL_SIZE) rehash(n / 2); return oldValue; } private double removeNullEntry() { containsNullKey = false; final double oldValue = value[n]; size--; if (n > minN && size < maxFill / 4 && n > DEFAULT_INITIAL_SIZE) rehash(n / 2); return oldValue; } @Override public void putAll(Map m) { if (f <= .5) ensureCapacity(m.size()); // The resulting map will be sized for m.size() elements else tryCapacity(size() + m.size()); // The resulting map will be tentatively sized for size() + m.size() // elements super.putAll(m); } private int find(final double k) { if ((Double.doubleToLongBits(k) == 0)) return containsNullKey ? n : -(n + 1); double curr; final double[] key = this.key; int pos; // The starting point. if ((Double.doubleToLongBits( curr = key[pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask]) == 0)) return -(pos + 1); if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return pos; // There's always an unused entry. while (true) { if ((Double.doubleToLongBits(curr = key[pos = (pos + 1) & mask]) == 0)) return -(pos + 1); if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return pos; } } private void insert(final int pos, final double k, final double v) { if (pos == n) containsNullKey = true; key[pos] = k; value[pos] = v; if (size++ >= maxFill) rehash(arraySize(size + 1, f)); if (ASSERTS) checkTable(); } @Override public double put(final double k, final double v) { final int pos = find(k); if (pos < 0) { insert(-pos - 1, k, v); return defRetValue; } final double oldValue = value[pos]; value[pos] = v; return oldValue; } private double addToValue(final int pos, final double incr) { final double oldValue = value[pos]; value[pos] = oldValue + incr; return oldValue; } /** * Adds an increment to value currently associated with a key. * *

* Note that this method respects the {@linkplain #defaultReturnValue() default * return value} semantics: when called with a key that does not currently * appears in the map, the key will be associated with the default return value * plus the given increment. * * @param k * the key. * @param incr * the increment. * @return the old value, or the {@linkplain #defaultReturnValue() default * return value} if no value was present for the given key. */ public double addTo(final double k, final double incr) { int pos; if ((Double.doubleToLongBits(k) == 0)) { if (containsNullKey) return addToValue(n, incr); pos = n; containsNullKey = true; } else { double curr; final double[] key = this.key; // The starting point. if (!(Double.doubleToLongBits( curr = key[pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask]) == 0)) { if ((Double.doubleToLongBits(curr) == Double.doubleToLongBits(k))) return addToValue(pos, incr); while (!(Double.doubleToLongBits(curr = key[pos = (pos + 1) & mask]) == 0)) if ((Double.doubleToLongBits(curr) == Double.doubleToLongBits(k))) return addToValue(pos, incr); } } key[pos] = k; value[pos] = defRetValue + incr; if (size++ >= maxFill) rehash(arraySize(size + 1, f)); if (ASSERTS) checkTable(); return defRetValue; } /** * 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(int pos) { // Shift entries with the same hash. int last, slot; double curr; final double[] key = this.key; for (;;) { pos = ((last = pos) + 1) & mask; for (;;) { if ((Double.doubleToLongBits(curr = key[pos]) == 0)) { key[last] = (0); return; } slot = (int) 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; } key[last] = curr; value[last] = value[pos]; } } @Override public double remove(final double k) { if ((Double.doubleToLongBits(k) == 0)) { if (containsNullKey) return removeNullEntry(); return defRetValue; } double curr; final double[] key = this.key; int pos; // The starting point. if ((Double.doubleToLongBits( curr = key[pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask]) == 0)) return defRetValue; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return removeEntry(pos); while (true) { if ((Double.doubleToLongBits(curr = key[pos = (pos + 1) & mask]) == 0)) return defRetValue; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return removeEntry(pos); } } @Override public double get(final double k) { if ((Double.doubleToLongBits(k) == 0)) return containsNullKey ? value[n] : defRetValue; double curr; final double[] key = this.key; int pos; // The starting point. if ((Double.doubleToLongBits( curr = key[pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask]) == 0)) return defRetValue; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return value[pos]; // There's always an unused entry. while (true) { if ((Double.doubleToLongBits(curr = key[pos = (pos + 1) & mask]) == 0)) return defRetValue; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return value[pos]; } } @Override public boolean containsKey(final double k) { if ((Double.doubleToLongBits(k) == 0)) return containsNullKey; double curr; final double[] key = this.key; int pos; // The starting point. if ((Double.doubleToLongBits( curr = key[pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask]) == 0)) return false; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return true; // There's always an unused entry. while (true) { if ((Double.doubleToLongBits(curr = key[pos = (pos + 1) & mask]) == 0)) return false; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return true; } } @Override public boolean containsValue(final double v) { final double value[] = this.value; final double key[] = this.key; if (containsNullKey && (Double.doubleToLongBits(value[n]) == Double.doubleToLongBits(v))) return true; for (int i = n; i-- != 0;) if (!(Double.doubleToLongBits(key[i]) == 0) && (Double.doubleToLongBits(value[i]) == Double.doubleToLongBits(v))) return true; return false; } /** {@inheritDoc} */ @Override public double getOrDefault(final double k, final double defaultValue) { if ((Double.doubleToLongBits(k) == 0)) return containsNullKey ? value[n] : defaultValue; double curr; final double[] key = this.key; int pos; // The starting point. if ((Double.doubleToLongBits( curr = key[pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask]) == 0)) return defaultValue; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return value[pos]; // There's always an unused entry. while (true) { if ((Double.doubleToLongBits(curr = key[pos = (pos + 1) & mask]) == 0)) return defaultValue; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return value[pos]; } } /** {@inheritDoc} */ @Override public double putIfAbsent(final double k, final double v) { final int pos = find(k); if (pos >= 0) return value[pos]; insert(-pos - 1, k, v); return defRetValue; } /** {@inheritDoc} */ @Override public boolean remove(final double k, final double v) { if ((Double.doubleToLongBits(k) == 0)) { if (containsNullKey && (Double.doubleToLongBits(v) == Double.doubleToLongBits(value[n]))) { removeNullEntry(); return true; } return false; } double curr; final double[] key = this.key; int pos; // The starting point. if ((Double.doubleToLongBits( curr = key[pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask]) == 0)) return false; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr)) && (Double.doubleToLongBits(v) == Double.doubleToLongBits(value[pos]))) { removeEntry(pos); return true; } while (true) { if ((Double.doubleToLongBits(curr = key[pos = (pos + 1) & mask]) == 0)) return false; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr)) && (Double.doubleToLongBits(v) == Double.doubleToLongBits(value[pos]))) { removeEntry(pos); return true; } } } /** {@inheritDoc} */ @Override public boolean replace(final double k, final double oldValue, final double v) { final int pos = find(k); if (pos < 0 || !(Double.doubleToLongBits(oldValue) == Double.doubleToLongBits(value[pos]))) return false; value[pos] = v; return true; } /** {@inheritDoc} */ @Override public double replace(final double k, final double v) { final int pos = find(k); if (pos < 0) return defRetValue; final double oldValue = value[pos]; value[pos] = v; return oldValue; } /** {@inheritDoc} */ @Override public double computeIfAbsent(final double k, final java.util.function.DoubleUnaryOperator mappingFunction) { java.util.Objects.requireNonNull(mappingFunction); final int pos = find(k); if (pos >= 0) return value[pos]; final double newValue = mappingFunction.applyAsDouble(k); insert(-pos - 1, k, newValue); return newValue; } /** {@inheritDoc} */ @Override public double computeIfAbsent(final double key, final Double2DoubleFunction mappingFunction) { java.util.Objects.requireNonNull(mappingFunction); final int pos = find(key); if (pos >= 0) return value[pos]; if (!mappingFunction.containsKey(key)) return defRetValue; final double newValue = mappingFunction.get(key); insert(-pos - 1, key, newValue); return newValue; } /** {@inheritDoc} */ @Override public double computeIfAbsentNullable(final double k, final java.util.function.DoubleFunction mappingFunction) { java.util.Objects.requireNonNull(mappingFunction); final int pos = find(k); if (pos >= 0) return value[pos]; final Double newValue = mappingFunction.apply(k); if (newValue == null) return defRetValue; final double v = (newValue).doubleValue(); insert(-pos - 1, k, v); return v; } /** {@inheritDoc} */ @Override public double computeIfPresent(final double k, final java.util.function.BiFunction remappingFunction) { java.util.Objects.requireNonNull(remappingFunction); final int pos = find(k); if (pos < 0) return defRetValue; final Double newValue = remappingFunction.apply(Double.valueOf(k), Double.valueOf(value[pos])); if (newValue == null) { if ((Double.doubleToLongBits(k) == 0)) removeNullEntry(); else removeEntry(pos); return defRetValue; } return value[pos] = (newValue).doubleValue(); } /** {@inheritDoc} */ @Override public double compute(final double k, final java.util.function.BiFunction remappingFunction) { java.util.Objects.requireNonNull(remappingFunction); final int pos = find(k); final Double newValue = remappingFunction.apply(Double.valueOf(k), pos >= 0 ? Double.valueOf(value[pos]) : null); if (newValue == null) { if (pos >= 0) { if ((Double.doubleToLongBits(k) == 0)) removeNullEntry(); else removeEntry(pos); } return defRetValue; } double newVal = (newValue).doubleValue(); if (pos < 0) { insert(-pos - 1, k, newVal); return newVal; } return value[pos] = newVal; } /** {@inheritDoc} */ @Override public double merge(final double k, final double v, final java.util.function.BiFunction remappingFunction) { java.util.Objects.requireNonNull(remappingFunction); final int pos = find(k); if (pos < 0) { insert(-pos - 1, k, v); return v; } final Double newValue = remappingFunction.apply(Double.valueOf(value[pos]), Double.valueOf(v)); if (newValue == null) { if ((Double.doubleToLongBits(k) == 0)) removeNullEntry(); else removeEntry(pos); return defRetValue; } return value[pos] = (newValue).doubleValue(); } /* * Removes all elements from this map. * *

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()}. * */ @Override public void clear() { if (size == 0) return; size = 0; containsNullKey = false; Arrays.fill(key, (0)); } @Override public int size() { return size; } @Override public boolean isEmpty() { return size == 0; } /** * The entry class for a hash map does not record key and value, but rather the * position in the hash table of the corresponding entry. This is necessary so * that calls to {@link java.util.Map.Entry#setValue(Object)} are reflected in * the map */ final class MapEntry implements Double2DoubleMap.Entry, Map.Entry, DoubleDoublePair { // The table index this entry refers to, or -1 if this entry has been deleted. int index; MapEntry(final int index) { this.index = index; } MapEntry() { } @Override public double getDoubleKey() { return key[index]; } @Override public double leftDouble() { return key[index]; } @Override public double getDoubleValue() { return value[index]; } @Override public double rightDouble() { return value[index]; } @Override public double setValue(final double v) { final double oldValue = value[index]; value[index] = v; return oldValue; } @Override public DoubleDoublePair right(final double v) { value[index] = v; return this; } /** * {@inheritDoc} * * @deprecated Please use the corresponding type-specific method instead. */ @Deprecated @Override public Double getKey() { return Double.valueOf(key[index]); } /** * {@inheritDoc} * * @deprecated Please use the corresponding type-specific method instead. */ @Deprecated @Override public Double getValue() { return Double.valueOf(value[index]); } /** * {@inheritDoc} * * @deprecated Please use the corresponding type-specific method instead. */ @Deprecated @Override public Double setValue(final Double v) { return Double.valueOf(setValue((v).doubleValue())); } @SuppressWarnings("unchecked") @Override public boolean equals(final Object o) { if (!(o instanceof Map.Entry)) return false; Map.Entry e = (Map.Entry) o; return (Double.doubleToLongBits(key[index]) == Double.doubleToLongBits((e.getKey()).doubleValue())) && (Double.doubleToLongBits(value[index]) == Double.doubleToLongBits((e.getValue()).doubleValue())); } @Override public int hashCode() { return it.unimi.dsi.fastutil.HashCommon.double2int(key[index]) ^ it.unimi.dsi.fastutil.HashCommon.double2int(value[index]); } @Override public String toString() { return key[index] + "=>" + value[index]; } } /** An iterator over a hash map. */ private abstract class MapIterator { /** * The index of the last entry returned, if positive or zero; initially, * {@link #n}. If negative, the last entry returned was that of the key of index * {@code - pos - 1} from the {@link #wrapped} list. */ int pos = n; /** * The index of the last entry that has been returned (more precisely, the value * of {@link #pos} if {@link #pos} is positive, or {@link Integer#MIN_VALUE} if * {@link #pos} is negative). It is -1 if either we did not return an entry yet, * or the last returned entry has been removed. */ int last = -1; /** A downward counter measuring how many entries must still be returned. */ int c = size; /** * A boolean telling us whether we should return the entry with the null key. */ boolean mustReturnNullKey = Double2DoubleOpenHashMap.this.containsNullKey; /** * A lazily allocated list containing keys of entries that have wrapped around * the table because of removals. */ DoubleArrayList wrapped; @SuppressWarnings("unused") abstract void acceptOnIndex(final ConsumerType action, final int index); public boolean hasNext() { return c != 0; } public int nextEntry() { if (!hasNext()) throw new NoSuchElementException(); c--; if (mustReturnNullKey) { mustReturnNullKey = false; return last = n; } final double key[] = Double2DoubleOpenHashMap.this.key; for (;;) { if (--pos < 0) { // We are just enumerating elements from the wrapped list. last = Integer.MIN_VALUE; final double k = wrapped.getDouble(-pos - 1); int p = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask; while (!(Double.doubleToLongBits(k) == Double.doubleToLongBits(key[p]))) p = (p + 1) & mask; return p; } if (!(Double.doubleToLongBits(key[pos]) == 0)) return last = pos; } } public void forEachRemaining(final ConsumerType action) { if (mustReturnNullKey) { mustReturnNullKey = false; acceptOnIndex(action, last = n); c--; } final double key[] = Double2DoubleOpenHashMap.this.key; while (c != 0) { if (--pos < 0) { // We are just enumerating elements from the wrapped list. last = Integer.MIN_VALUE; final double k = wrapped.getDouble(-pos - 1); int p = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask; while (!(Double.doubleToLongBits(k) == Double.doubleToLongBits(key[p]))) p = (p + 1) & mask; acceptOnIndex(action, p); c--; } else if (!(Double.doubleToLongBits(key[pos]) == 0)) { acceptOnIndex(action, last = pos); c--; } } } /** * 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 void shiftKeys(int pos) { // Shift entries with the same hash. int last, slot; double curr; final double[] key = Double2DoubleOpenHashMap.this.key; for (;;) { pos = ((last = pos) + 1) & mask; for (;;) { if ((Double.doubleToLongBits(curr = key[pos]) == 0)) { key[last] = (0); return; } slot = (int) 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(2); wrapped.add(key[pos]); } key[last] = curr; value[last] = value[pos]; } } public void remove() { if (last == -1) throw new IllegalStateException(); if (last == n) { containsNullKey = false; } else if (pos >= 0) shiftKeys(last); else { // We're removing wrapped entries. Double2DoubleOpenHashMap.this.remove(wrapped.getDouble(-pos - 1)); last = -1; // Note that we must not decrement size return; } size--; last = -1; // You can no longer remove this entry. if (ASSERTS) checkTable(); } public int skip(final int n) { int i = n; while (i-- != 0 && hasNext()) nextEntry(); return n - i - 1; } } private final class EntryIterator extends MapIterator> implements ObjectIterator { private MapEntry entry; @Override public MapEntry next() { return entry = new MapEntry(nextEntry()); } // forEachRemaining inherited from MapIterator superclass. @Override final void acceptOnIndex(final Consumer action, final int index) { action.accept(entry = new MapEntry(index)); } @Override public void remove() { super.remove(); entry.index = -1; // You cannot use a deleted entry. } } private final class FastEntryIterator extends MapIterator> implements ObjectIterator { private final MapEntry entry = new MapEntry(); @Override public MapEntry next() { entry.index = nextEntry(); return entry; } // forEachRemaining inherited from MapIterator superclass. @Override final void acceptOnIndex(final Consumer action, final int index) { entry.index = index; action.accept(entry); } } private abstract class MapSpliterator> { /** * The index (which bucket) of the next item to give to the action. Unlike * {@link SetIterator}, this counts up instead of down. */ int pos = 0; /** The maximum bucket (exclusive) to iterate to */ int max = n; /** An upwards counter counting how many we have given */ int c = 0; /** A boolean telling us whether we should return the null key. */ boolean mustReturnNull = Double2DoubleOpenHashMap.this.containsNullKey; boolean hasSplit = false; MapSpliterator() { } MapSpliterator(int pos, int max, boolean mustReturnNull, boolean hasSplit) { this.pos = pos; this.max = max; this.mustReturnNull = mustReturnNull; this.hasSplit = hasSplit; } abstract void acceptOnIndex(final ConsumerType action, final int index); abstract SplitType makeForSplit(int pos, int max, boolean mustReturnNull); public boolean tryAdvance(final ConsumerType action) { if (mustReturnNull) { mustReturnNull = false; ++c; acceptOnIndex(action, n); return true; } final double key[] = Double2DoubleOpenHashMap.this.key; while (pos < max) { if (!(Double.doubleToLongBits(key[pos]) == 0)) { ++c; acceptOnIndex(action, pos++); return true; } ++pos; } return false; } public void forEachRemaining(final ConsumerType action) { if (mustReturnNull) { mustReturnNull = false; ++c; acceptOnIndex(action, n); } final double key[] = Double2DoubleOpenHashMap.this.key; while (pos < max) { if (!(Double.doubleToLongBits(key[pos]) == 0)) { acceptOnIndex(action, pos); ++c; } ++pos; } } public long estimateSize() { if (!hasSplit) { // Root spliterator; we know how many are remaining. return size - c; } else { // After we split, we can no longer know exactly how many we have (or at least // not efficiently). // (size / n) * (max - pos) aka currentTableDensity * numberOfBucketsLeft seems // like a good estimate. return Math.min(size - c, (long) (((double) realSize() / n) * (max - pos)) + (mustReturnNull ? 1 : 0)); } } public SplitType trySplit() { if (pos >= max - 1) return null; int retLen = (max - pos) >> 1; if (retLen <= 1) return null; int myNewPos = pos + retLen; int retPos = pos; int retMax = myNewPos; // Since null is returned first, and the convention is that the returned split // is the prefix of elements, // the split will take care of returning null (if needed), and we won't return // it anymore. SplitType split = makeForSplit(retPos, retMax, mustReturnNull); this.pos = myNewPos; this.mustReturnNull = false; this.hasSplit = true; return split; } public long skip(long n) { if (n < 0) throw new IllegalArgumentException("Argument must be nonnegative: " + n); if (n == 0) return 0; long skipped = 0; if (mustReturnNull) { mustReturnNull = false; ++skipped; --n; } final double key[] = Double2DoubleOpenHashMap.this.key; while (pos < max && n > 0) { if (!(Double.doubleToLongBits(key[pos++]) == 0)) { ++skipped; --n; } } return skipped; } } private final class EntrySpliterator extends MapSpliterator, EntrySpliterator> implements ObjectSpliterator { private static final int POST_SPLIT_CHARACTERISTICS = ObjectSpliterators.SET_SPLITERATOR_CHARACTERISTICS & ~java.util.Spliterator.SIZED; EntrySpliterator() { } EntrySpliterator(int pos, int max, boolean mustReturnNull, boolean hasSplit) { super(pos, max, mustReturnNull, hasSplit); } @Override public int characteristics() { return hasSplit ? POST_SPLIT_CHARACTERISTICS : ObjectSpliterators.SET_SPLITERATOR_CHARACTERISTICS; } @Override final void acceptOnIndex(final Consumer action, final int index) { action.accept(new MapEntry(index)); } @Override final EntrySpliterator makeForSplit(int pos, int max, boolean mustReturnNull) { return new EntrySpliterator(pos, max, mustReturnNull, true); } } private final class MapEntrySet extends AbstractObjectSet implements FastEntrySet { @Override public ObjectIterator iterator() { return new EntryIterator(); } @Override public ObjectIterator fastIterator() { return new FastEntryIterator(); } @Override public ObjectSpliterator spliterator() { return new EntrySpliterator(); } // @Override public boolean contains(final Object o) { if (!(o instanceof Map.Entry)) return false; final Map.Entry e = (Map.Entry) o; if (e.getKey() == null || !(e.getKey() instanceof Double)) return false; if (e.getValue() == null || !(e.getValue() instanceof Double)) return false; final double k = ((Double) (e.getKey())).doubleValue(); final double v = ((Double) (e.getValue())).doubleValue(); if ((Double.doubleToLongBits(k) == 0)) return Double2DoubleOpenHashMap.this.containsNullKey && (Double.doubleToLongBits(value[n]) == Double.doubleToLongBits(v)); double curr; final double[] key = Double2DoubleOpenHashMap.this.key; int pos; // The starting point. if ((Double.doubleToLongBits( curr = key[pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask]) == 0)) return false; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return (Double.doubleToLongBits(value[pos]) == Double.doubleToLongBits(v)); // There's always an unused entry. while (true) { if ((Double.doubleToLongBits(curr = key[pos = (pos + 1) & mask]) == 0)) return false; if ((Double.doubleToLongBits(k) == Double.doubleToLongBits(curr))) return (Double.doubleToLongBits(value[pos]) == Double.doubleToLongBits(v)); } } @Override public boolean remove(final Object o) { if (!(o instanceof Map.Entry)) return false; final Map.Entry e = (Map.Entry) o; if (e.getKey() == null || !(e.getKey() instanceof Double)) return false; if (e.getValue() == null || !(e.getValue() instanceof Double)) return false; final double k = ((Double) (e.getKey())).doubleValue(); final double v = ((Double) (e.getValue())).doubleValue(); if ((Double.doubleToLongBits(k) == 0)) { if (containsNullKey && (Double.doubleToLongBits(value[n]) == Double.doubleToLongBits(v))) { removeNullEntry(); return true; } return false; } double curr; final double[] key = Double2DoubleOpenHashMap.this.key; int pos; // The starting point. if ((Double.doubleToLongBits( curr = key[pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask]) == 0)) return false; if ((Double.doubleToLongBits(curr) == Double.doubleToLongBits(k))) { if ((Double.doubleToLongBits(value[pos]) == Double.doubleToLongBits(v))) { removeEntry(pos); return true; } return false; } while (true) { if ((Double.doubleToLongBits(curr = key[pos = (pos + 1) & mask]) == 0)) return false; if ((Double.doubleToLongBits(curr) == Double.doubleToLongBits(k))) { if ((Double.doubleToLongBits(value[pos]) == Double.doubleToLongBits(v))) { removeEntry(pos); return true; } } } } @Override public int size() { return size; } @Override public void clear() { Double2DoubleOpenHashMap.this.clear(); } /** {@inheritDoc} */ @Override public void forEach(final Consumer consumer) { if (containsNullKey) consumer.accept(new AbstractDouble2DoubleMap.BasicEntry(key[n], value[n])); for (int pos = n; pos-- != 0;) if (!(Double.doubleToLongBits(key[pos]) == 0)) consumer.accept(new AbstractDouble2DoubleMap.BasicEntry(key[pos], value[pos])); } /** {@inheritDoc} */ @Override public void fastForEach(final Consumer consumer) { final AbstractDouble2DoubleMap.BasicEntry entry = new AbstractDouble2DoubleMap.BasicEntry(); if (containsNullKey) { entry.key = key[n]; entry.value = value[n]; consumer.accept(entry); } for (int pos = n; pos-- != 0;) if (!(Double.doubleToLongBits(key[pos]) == 0)) { entry.key = key[pos]; entry.value = value[pos]; consumer.accept(entry); } } } @Override public FastEntrySet double2DoubleEntrySet() { if (entries == null) entries = new MapEntrySet(); return entries; } /** * An iterator on keys. * *

* We simply override the * {@link java.util.ListIterator#next()}/{@link java.util.ListIterator#previous()} * methods (and possibly their type-specific counterparts) so that they return * keys instead of entries. */ private final class KeyIterator extends MapIterator implements DoubleIterator { public KeyIterator() { super(); } // forEachRemaining inherited from MapIterator superclass. // Despite the superclass declared with generics, the way Java inherits and // generates bridge methods avoids the boxing/unboxing @Override final void acceptOnIndex(final java.util.function.DoubleConsumer action, final int index) { action.accept(key[index]); } @Override public double nextDouble() { return key[nextEntry()]; } } private final class KeySpliterator extends MapSpliterator implements DoubleSpliterator { private static final int POST_SPLIT_CHARACTERISTICS = DoubleSpliterators.SET_SPLITERATOR_CHARACTERISTICS & ~java.util.Spliterator.SIZED; KeySpliterator() { } KeySpliterator(int pos, int max, boolean mustReturnNull, boolean hasSplit) { super(pos, max, mustReturnNull, hasSplit); } @Override public int characteristics() { return hasSplit ? POST_SPLIT_CHARACTERISTICS : DoubleSpliterators.SET_SPLITERATOR_CHARACTERISTICS; } @Override final void acceptOnIndex(final java.util.function.DoubleConsumer action, final int index) { action.accept(key[index]); } @Override final KeySpliterator makeForSplit(int pos, int max, boolean mustReturnNull) { return new KeySpliterator(pos, max, mustReturnNull, true); } } private final class KeySet extends AbstractDoubleSet { @Override public DoubleIterator iterator() { return new KeyIterator(); } @Override public DoubleSpliterator spliterator() { return new KeySpliterator(); } /** {@inheritDoc} */ @Override public void forEach(final java.util.function.DoubleConsumer consumer) { if (containsNullKey) consumer.accept(key[n]); for (int pos = n; pos-- != 0;) { final double k = key[pos]; if (!(Double.doubleToLongBits(k) == 0)) consumer.accept(k); } } @Override public int size() { return size; } @Override public boolean contains(double k) { return containsKey(k); } @Override public boolean remove(double k) { final int oldSize = size; Double2DoubleOpenHashMap.this.remove(k); return size != oldSize; } @Override public void clear() { Double2DoubleOpenHashMap.this.clear(); } } @Override public DoubleSet keySet() { if (keys == null) keys = new KeySet(); return keys; } /** * An iterator on values. * *

* We simply override the * {@link java.util.ListIterator#next()}/{@link java.util.ListIterator#previous()} * methods (and possibly their type-specific counterparts) so that they return * values instead of entries. */ private final class ValueIterator extends MapIterator implements DoubleIterator { public ValueIterator() { super(); } // forEachRemaining inherited from MapIterator superclass. // Despite the superclass declared with generics, the way Java inherits and // generates bridge methods avoids the boxing/unboxing @Override final void acceptOnIndex(final java.util.function.DoubleConsumer action, final int index) { action.accept(value[index]); } @Override public double nextDouble() { return value[nextEntry()]; } } private final class ValueSpliterator extends MapSpliterator implements DoubleSpliterator { private static final int POST_SPLIT_CHARACTERISTICS = DoubleSpliterators.COLLECTION_SPLITERATOR_CHARACTERISTICS & ~java.util.Spliterator.SIZED; ValueSpliterator() { } ValueSpliterator(int pos, int max, boolean mustReturnNull, boolean hasSplit) { super(pos, max, mustReturnNull, hasSplit); } @Override public int characteristics() { return hasSplit ? POST_SPLIT_CHARACTERISTICS : DoubleSpliterators.COLLECTION_SPLITERATOR_CHARACTERISTICS; } @Override final void acceptOnIndex(final java.util.function.DoubleConsumer action, final int index) { action.accept(value[index]); } @Override final ValueSpliterator makeForSplit(int pos, int max, boolean mustReturnNull) { return new ValueSpliterator(pos, max, mustReturnNull, true); } } @Override public DoubleCollection values() { if (values == null) values = new AbstractDoubleCollection() { @Override public DoubleIterator iterator() { return new ValueIterator(); } @Override public DoubleSpliterator spliterator() { return new ValueSpliterator(); } /** {@inheritDoc} */ @Override public void forEach(final java.util.function.DoubleConsumer consumer) { if (containsNullKey) consumer.accept(value[n]); for (int pos = n; pos-- != 0;) if (!(Double.doubleToLongBits(key[pos]) == 0)) consumer.accept(value[pos]); } @Override public int size() { return size; } @Override public boolean contains(double v) { return containsValue(v); } @Override public void clear() { Double2DoubleOpenHashMap.this.clear(); } }; return values; } /** * Rehashes the map, 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 map. * @see #trim(int) */ public boolean trim() { return trim(size); } /** * Rehashes this map if the table is too large. * *

* Let N be the smallest table size that can hold * max(n,{@link #size()}) 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 map in a table of size * N. * *

* This method is useful when reusing maps. {@linkplain #clear() Clearing a map} * leaves the table size untouched. If you are reusing a map 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 maps. * * @param n * the threshold for the trimming. * @return true if there was enough memory to trim the map. * @see #trim() */ public boolean trim(final int n) { final int l = HashCommon.nextPowerOfTwo((int) Math.ceil(n / f)); if (l >= this.n || size > maxFill(l, f)) return true; try { rehash(l); } catch (OutOfMemoryError cantDoIt) { return false; } return true; } /** * Rehashes the map. * *

* This method implements the basic rehashing strategy, and may be overridden 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 int newN) { final double key[] = this.key; final double value[] = this.value; final int mask = newN - 1; // Note that this is used by the hashing macro final double newKey[] = new double[newN + 1]; final double newValue[] = new double[newN + 1]; int i = n, pos; for (int j = realSize(); j-- != 0;) { while ((Double.doubleToLongBits(key[--i]) == 0)); if (!(Double.doubleToLongBits( newKey[pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(key[i])) & mask]) == 0)) while (!(Double.doubleToLongBits(newKey[pos = (pos + 1) & mask]) == 0)); newKey[pos] = key[i]; newValue[pos] = value[i]; } newValue[newN] = value[n]; n = newN; this.mask = mask; maxFill = maxFill(n, f); this.key = newKey; this.value = newValue; } /** * Returns a deep copy of this map. * *

* This method performs a deep copy of this hash map; the data stored in the * map, however, is not cloned. Note that this makes a difference only for * object keys. * * @return a deep copy of this map. */ @Override public Double2DoubleOpenHashMap clone() { Double2DoubleOpenHashMap c; try { c = (Double2DoubleOpenHashMap) super.clone(); } catch (CloneNotSupportedException cantHappen) { throw new InternalError(); } c.keys = null; c.values = null; c.entries = null; c.containsNullKey = containsNullKey; c.key = key.clone(); c.value = value.clone(); return c; } /** * Returns a hash code for this map. * * 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 map. */ @Override public int hashCode() { int h = 0; for (int j = realSize(), i = 0, t = 0; j-- != 0;) { while ((Double.doubleToLongBits(key[i]) == 0)) i++; t = it.unimi.dsi.fastutil.HashCommon.double2int(key[i]); t ^= it.unimi.dsi.fastutil.HashCommon.double2int(value[i]); h += t; i++; } // Zero / null keys have hash zero. if (containsNullKey) h += it.unimi.dsi.fastutil.HashCommon.double2int(value[n]); return h; } private void writeObject(java.io.ObjectOutputStream s) throws java.io.IOException { final double key[] = this.key; final double value[] = this.value; final EntryIterator i = new EntryIterator(); s.defaultWriteObject(); for (int j = size, e; j-- != 0;) { e = i.nextEntry(); s.writeDouble(key[e]); s.writeDouble(value[e]); } } private void readObject(java.io.ObjectInputStream s) throws java.io.IOException, ClassNotFoundException { s.defaultReadObject(); n = arraySize(size, f); maxFill = maxFill(n, f); mask = n - 1; final double key[] = this.key = new double[n + 1]; final double value[] = this.value = new double[n + 1]; double k; double v; for (int i = size, pos; i-- != 0;) { k = s.readDouble(); v = s.readDouble(); if ((Double.doubleToLongBits(k) == 0)) { pos = n; containsNullKey = true; } else { pos = (int) it.unimi.dsi.fastutil.HashCommon.mix(Double.doubleToRawLongBits(k)) & mask; while (!(Double.doubleToLongBits(key[pos]) == 0)) pos = (pos + 1) & mask; } key[pos] = k; value[pos] = v; } if (ASSERTS) checkTable(); } private void checkTable() { } }





© 2015 - 2024 Weber Informatics LLC | Privacy Policy