转载请注明出处:http://blog.csdn.net/zhoubin1992/article/details/48269985
前言
当我们定义
HashMap<Integer, E> hashMap = new HashMap<Integer, E>(); 时IDE会给出一个 警告:
用SparseArray<E>来替代,以获取更好性能。HashMap的get和put时间复杂度是O(1)呀,为什么SparseArray的性能会更好?这两天花时间详细分析了下源码,发现SparseArray的优点主要是节约内存,非常适合于移动端。
SparseArray是在android.util下的一个工具类。首先我们来看下SparseArray的方法结构:
会发现内容并不多,就是构造方法和增删改查的一些方法。
源码分析
源码详细分析见注释:
package android.util;
import com.android.internal.util.ArrayUtils;
import com.android.internal.util.GrowingArrayUtils;
import libcore.util.EmptyArray;
/**
* SparseArrays map integers to Objects. Unlike a normal array of Objects,
* there can be gaps in the indices. It is intended to be more memory efficient
* than using a HashMap to map Integers to Objects, both because it avoids
* auto-boxing keys and its data structure doesn't rely on an extra entry object
* for each mapping.
* SparseArray用于映射integers到object。但不像普通数组那样,sparseArray的元素间没有无用元素。
* 在映射integers到object的过程中,SparseArray由于采用避免自动装箱的keys和它的数据结构不依赖额外
* 的对象来存储映射关系的实现,因此它比hashMap的内存使用更高效一些。
* <p>Note that this container keeps its mappings in an array data structure,
* using a binary search to find keys. The implementation is not intended to be appropriate for
* data structures
* that may contain large numbers of items. It is generally slower than a traditional
* HashMap, since lookups require a binary search and adds and removes require inserting
* and deleting entries in the array. For containers holding up to hundreds of items,
* the performance difference is not significant, less than 50%.</p>
* 注意:SparseArray在查找keys的过程中采用了二分查找, 这种实现不适合数据量大的情况。由于查找时要用到二分查找,
* 添加删除时涉及到数组其他元素的挪动,因此通常SparseArray会比hashMap慢。当处理上百的数据量,这种性能差异不是特别
* 明显,性能差异不超过50%。
* <p>To help with performance, the container includes an optimization when removing
* keys: instead of compacting its array immediately, it leaves the removed entry marked
* as deleted. The entry can then be re-used for the same key, or compacted later in
* a single garbage collection step of all removed entries. This garbage collection will
* need to be performed at any time the array needs to be grown or the the map size or
* entry values are retrieved.</p>
*为了优化性能,SparseArray针对remove case作了优化,remove时它不是立即挤压数组空间,而是标记为delete。
* 这个被标记的元素要么被重复利用,要么在多次remove之后通过一次gc操作中被挤压出去。
* gc需要在下列情况之前被执行:数组要扩容;获取SparseArray容量;get values(更详细的见代码注释);
* <p>It is possible to iterate over the items in this container using
* {@link #keyAt(int)} and {@link #valueAt(int)}. Iterating over the keys using
* <code>keyAt(int)</code> with ascending values of the index will return the
* keys in ascending order, or the values corresponding to the keys in ascending
* order in the case of <code>valueAt(int)</code>.</p>
* 可以用keyAt valueAt实现遍历....
*/
//E对应HashMap的Value
public class SparseArray<E> implements Cloneable {
// 用来优化删除性能(当有元素被remove delete时),标记已经删除的对象
private static final Object DELETED = new Object();
// 用来优化删除性能,标记是否需要垃圾回收
private boolean mGarbage = false;
// 存储索引,整数索引(key为整数)从小到大被映射在该数组
private int[] mKeys;
// 存储对象(Value)
private Object[] mValues;
// SparseArray实际大小
private int mSize;
/**
* Creates a new SparseArray containing no mappings.
*/
public SparseArray() {
//默认容量是10个元素
this(10);
}
/**
* Creates a new SparseArray containing no mappings that will not
* require any additional memory allocation to store the specified
* number of mappings. If you supply an initial capacity of 0, the
* sparse array will be initialized with a light-weight representation
* not requiring any additional array allocations.
*/
public SparseArray(int initialCapacity) {
if (initialCapacity == 0) {
//mKeys的初值等于new int[0],mValues的初值等于new Object[0]
mKeys = EmptyArray.INT;
mValues = EmptyArray.OBJECT;
} else {
//newUnpaddedObjectArray最后指向了VMRuntime的一个native方法,返回一个至少长initialCapacity的数组,
//但可能更大。增长的大小来自于避免数组后的任何padding。padding的大小依赖于componentType和内存分配器的实现。
mValues = ArrayUtils.newUnpaddedObjectArray(initialCapacity);
mKeys = new int[mValues.length];
}
mSize = 0;
}
@Override
@SuppressWarnings("unchecked")
public SparseArray<E> clone() {
SparseArray<E> clone = null;
try {
//java深拷贝
clone = (SparseArray<E>) super.clone();
clone.mKeys = mKeys.clone();
clone.mValues = mValues.clone();
} catch (CloneNotSupportedException cnse) {
/* ignore */
}
return clone;
}
/**
* Gets the Object mapped from the specified key, or <code>null</code>
* if no such mapping has been made.
*/
/**
* 获得指定key的映射对象,或者null如果没有该映射。
*/
public E get(int key) {
return get(key, null);
}
/**
* Gets the Object mapped from the specified key, or the specified Object
* if no such mapping has been made.
*/
@SuppressWarnings("unchecked")
public E get(int key, E valueIfKeyNotFound) {
//二分查找
int i = ContainerHelpers.binarySearch(mKeys, mSize, key);
// 如果没找到或者该value已经被标记删除,则返回默认值
if (i < 0 || mValues[i] == DELETED) {
return valueIfKeyNotFound;
} else {
// i>0 且该位置的元素未被标记为待删除,返回该值mValues[i]
return (E) mValues[i];
}
}
/**
* Removes the mapping from the specified key, if there was any.
*/
/**
* 删除指定key的映射对象。
*/
public void delete(int key) {
//二分查找
int i = ContainerHelpers.binarySearch(mKeys, mSize, key);
//找到了
if (i >= 0) {
//若未被标记delete,标记为delete,回收mGarbage=true
if (mValues[i] != DELETED) {
mValues[i] = DELETED;
mGarbage = true;
}
}
}
/**
* Alias for {@link #delete(int)}.
*/
public void remove(int key) {
delete(key);
}
/**
* Removes the mapping at the specified index.
*/
//移除特定位置的元素,注意传入的是mValues的index不是Key
public void removeAt(int index) {
//若未被标记delete,标记为delete,回收mGarbage=true
if (mValues[index] != DELETED) {
mValues[index] = DELETED;
mGarbage = true;
}
}
/**
* Remove a range of mappings as a batch.
*
* @param index Index to begin at
* @param size Number of mappings to remove
*/
public void removeAtRange(int index, int size) {
//确定结束位置
final int end = Math.min(mSize, index + size);
//从起点开始循环 remove
for (int i = index; i < end; i++) {
removeAt(i);
}
}
//目的只有一个压缩空间(压缩数组,把无效的值删除)
private void gc() {
// Log.e("SparseArray", "gc start with " + mSize);
int n = mSize;
int o = 0;
int[] keys = mKeys;
Object[] values = mValues;
//循环整个元素区间,删除值为DELETED的数,这里比较巧妙,直接对同一个keys和values操作,完成元素的删除和移动!
//大家注意这里的keys和mKeys是指向同一个地址(values 和mValues类似),所以改变keys、values 的值也即改变了mKeys、mValues。
for (int i = 0; i < n; i++) {
Object val = values[i];
if (val != DELETED) {
if (i != o) {
keys[o] = keys[i];
values[o] = val;
values[i] = null;
}
o++;
}
}
mGarbage = false;
mSize = o;//实际大小
// Log.e("SparseArray", "gc end with " + mSize);
}
/**
* Adds a mapping from the specified key to the specified value,
* replacing the previous mapping from the specified key if there
* was one.
*/
/**
* 添加一个指定key到指定object的映射,如果之前有一个指定key的映射则直接替换掉原映射object。注意gc。
*/
public void put(int key, E value) {
//先二分查找,确定插入位置,保证了key数组的有序性
int i = ContainerHelpers.binarySearch(mKeys, mSize, key);
if (i >= 0) {
//找到了,直接替换
mValues[i] = value;
} else {
// 做一个取反运算,获得应该插入的index
//没找到的情况下: i = -insertPoint -1,对他取反刚好得insertPoint。
i = ~i;
//若i在size范围内,且刚好对应位置标记为delete了,直接放入
if (i < mSize && mValues[i] == DELETED) {
mKeys[i] = key;
mValues[i] = value;
return;
}
//若前面if不成立,即i超出了size范围,或者对应的位置的元素是有效的
// 如果被标记为需要垃圾回收且SparseArray大小不小于keys数组长度
if (mGarbage && mSize >= mKeys.length) {
// 压缩空间,会压缩数组,把无效的值都去掉,保证连续有效值
gc();
// Search again because indices may have changed.
// 再次查找插入点因为索引可能改变
i = ~ContainerHelpers.binarySearch(mKeys, mSize, key);
}
// 插入,如果size不够则会重新分配更大的数组,然后拷贝过去并插入;size足够则用System.arraycopy把插入位置开始的value都后移然后插入
mKeys = GrowingArrayUtils.insert(mKeys, mSize, i, key);
mValues = GrowingArrayUtils.insert(mValues, mSize, i, value);
// 实际大小加1
mSize++;
}
}
/**
* Returns the number of key-value mappings that this SparseArray
* currently stores.
*/
//返回mSize,注意gc。
public int size() {
if (mGarbage) {
gc();
}
return mSize;
}
/**
* Given an index in the range <code>0...size()-1</code>, returns
* the key from the <code>index</code>th key-value mapping that this
* SparseArray stores.
*
* <p>The keys corresponding to indices in ascending order are guaranteed to
* be in ascending order, e.g., <code>keyAt(0)</code> will return the
* smallest key and <code>keyAt(size()-1)</code> will return the largest
* key.</p>
*/
//返回索引为index的mKeys值,注意gc。
public int keyAt(int index) {
if (mGarbage) {
gc();
}
return mKeys[index];
}
/**
* Given an index in the range <code>0...size()-1</code>, returns
* the value from the <code>index</code>th key-value mapping that this
* SparseArray stores.
*
* <p>The values corresponding to indices in ascending order are guaranteed
* to be associated with keys in ascending order, e.g.,
* <code>valueAt(0)</code> will return the value associated with the
* smallest key and <code>valueAt(size()-1)</code> will return the value
* associated with the largest key.</p>
*/
@SuppressWarnings("unchecked")
//返回索引为index的mValues值,注意gc。
public E valueAt(int index) {
if (mGarbage) {
gc();
}
return (E) mValues[index];
}
/**
* Given an index in the range <code>0...size()-1</code>, sets a new
* value for the <code>index</code>th key-value mapping that this
* SparseArray stores.
*/
//设置索引为index的mValues值为value,注意gc。
public void setValueAt(int index, E value) {
if (mGarbage) {
gc();
}
mValues[index] = value;
}
/**
* Returns the index for which {@link #keyAt} would return the
* specified key, or a negative number if the specified
* key is not mapped.
*/
//返回值为key在mKeys中的index,注意gc。
public int indexOfKey(int key) {
if (mGarbage) {
gc();
}
return ContainerHelpers.binarySearch(mKeys, mSize, key);
}
/**
* Returns an index for which {@link #valueAt} would return the
* specified key, or a negative number if no keys map to the
* specified value.
* <p>Beware that this is a linear search, unlike lookups by key,
* and that multiple keys can map to the same value and this will
* find only one of them.
* <p>Note also that unlike most collections' {@code indexOf} methods,
* this method compares values using {@code ==} rather than {@code equals}.
*/
//返回值为value在mValues中的index,注意gc。
public int indexOfValue(E value) {
if (mGarbage) {
gc();
}
for (int i = 0; i < mSize; i++)
if (mValues[i] == value)
return i;
return -1;
}
/**
* Removes all key-value mappings from this SparseArray.
*/
//清空SparseArray
public void clear() {
int n = mSize;
Object[] values = mValues;
for (int i = 0; i < n; i++) {
//值空,利于jvm gc
values[i] = null;
}
mSize = 0;
mGarbage = false;
}
/**
* Puts a key/value pair into the array, optimizing for the case where
* the key is greater than all existing keys in the array.
*/
//往SparseArray加入键值对key/value
public void append(int key, E value) {
//若key小于等于已有的最大key,直接Put
if (mSize != 0 && key <= mKeys[mSize - 1]) {
put(key, value);
return;
}
if (mGarbage && mSize >= mKeys.length) {
gc();
}
//若key大于了现有的所有key,就不用走put的二分查找过程了,直接append
mKeys = GrowingArrayUtils.append(mKeys, mSize, key);
mValues = GrowingArrayUtils.append(mValues, mSize, value);
mSize++;
}
/**
* {@inheritDoc}
*
* <p>This implementation composes a string by iterating over its mappings. If
* this map contains itself as a value, the string "(this Map)"
* will appear in its place.
*/
@Override
public String toString() {
if (size() <= 0) {
return "{}";
}
StringBuilder buffer = new StringBuilder(mSize * 28);
buffer.append('{');
for (int i=0; i<mSize; i++) {
if (i > 0) {
buffer.append(", ");
}
int key = keyAt(i);
buffer.append(key);
buffer.append('=');
Object value = valueAt(i);
if (value != this) {
buffer.append(value);
} else {
buffer.append("(this Map)");
}
}
buffer.append('}');
return buffer.toString();
}
}
SparseArray性能体现
阅读上面代码,我来总结下SparseArray的性能体现:
1. SparseArray用于映射integers到object。因为key是一个整数数组,避免了自动装箱的keys和不依赖额外的数据结构去映射K/V关系,从而节省内存,比hashMap的内存使用更高效。
2. SparseArray在get、put中采用二分查找,添加删除元素时涉及到数组其他元素的挪动,因此通常SparseArray会比hashMap慢。当处理上百的数据量,这种性能差异不是特别明显,性能差异不超过50%。所以SparseArray不适合数据量大的情况。
二分查找函数(binarySearch)代码如下:
private static int binarySearch(int[] a, int start, int len, int key) {
int high = start + len, low = start - 1, guess;
while (high - low > 1) {
guess = (high + low) / 2;
if (a[guess] < key)
low = guess;
else
high = guess;
}
if (high == start + len)
return ~(start + len);
else if (a[high] == key)
return high;
else
return ~high;
}
返回值是插入位置,没找到的情况下: i = -insertPoint -1 <0,对他取反(i=~i)刚好得insertPoint。
3.由于key数组需要有序,所以每次的put操作更费时,要二分查找,要在数组删除插入元素。所以对应地,SparseArray针对remove作了优化,remove元素时不是立即压缩数组空间,而是把需要remove的元素标记为delete,同时设置垃圾回收标志mGarbage。这个被标记的元素要么被重复利用,要么在多次remove之后通过一次gc操作中被挤压出去。
关于SparseArray的gc方法
- 注意gc方法只在如下方法会执行,在remove和get时不会执行。这就是针对remove作了优化,将可能的多次gc操作变为一次完成,见下一条。
- 用删除标记DELETED结合垃圾回收标记mGarbage减少gc次数。如果每次remove都执行一次gc(循环数组删除移动),则性能下降。采用优化后,多次remove仅多次设置了标志,在gc触发时,仅需要一次循环就可以将空间压缩好。
- gc方法的目的只有一个:压缩空间(压缩数组,把无效的值删除)。查看代码会发现循环整个元素区间,删除值为DELETED的数,这里比较巧妙,直接对同一个keys和values操作,完成元素的删除和移动!同时注意这里的keys和mKeys是指向同一个地址(values 和mValues类似),所以改变keys、values 的值也即改变了mKeys、mValues。
- 每次gc过程,保证了他的数组(mSize)区间内没有无效值。也是稀疏数组的精髓。
总结
与HashMap比较
HashMap的相关基础和源码见 JAVA 面向对象和集合知识点总结(6.15更新)
HashMap的get put时间复杂度O(1)的代价就是耗费大量内存来存储数据。
HashMap :消耗内存空间,性能好。
SparseArray :节约内存空间,性能稍差(上百的数据量性能差异不超过50%)。
适用情况
key为整数,同时K/V对数量不是太大,可以考虑使用SparseArray能更节省内存并且性能损耗不大。所以对于移动端这种内存珍贵来说,非常适用。
在Android中,用SparseArray<E> sparseArray = new SparseArray<E>(); 来替换
HashMap<Integer, E> hashMap = new HashMap<Integer, E>()吧~
本文分析了Android中的SparseArray,作为HashMap的替代者,它通过避免自动装箱和使用二分查找节省内存。尽管在添加删除元素时较慢,但适用于小型数据集。SparseArray在remove时采用删除标记,通过gc方法批量压缩空间,减少GC次数。总结建议,当key为整数且数据量不大时,使用SparseArray可以节省内存。

1333

被折叠的 条评论
为什么被折叠?



