java性能优化常用技巧

一、避免在循环条件中使用复杂表达式 

在不做编译优化的情况下，在循环中，循环条件会被反复计算，如果不使用复杂表达式，而使循环条件值不变的话，程序将会运行的更快。 

例子： 

import java.util.vector;
class cel {
    void method (vector vector) {
        for (int i = 0; i < vector.size (); i++)  // violation
            ; // ...
    }
}

import java.util.vector;
class cel {
    void method (vector vector) {
        for (int i = 0; i < vector.size (); i++)  // violation
            ; // ...
    }
}

更正： 

class cel_fixed {
    void method (vector vector) {
        int size = vector.size ()
        for (int i = 0; i < size; i++)
            ; // ...
    }
}

class cel_fixed {
    void method (vector vector) {
        int size = vector.size ()
        for (int i = 0; i < size; i++)
            ; // ...
    }
}

二、为'vectors' 和 'hashtables'定义初始大小 

jvm为vector扩充大小的时候需要重新创建一个更大的数组，将原原先数组中的内容复制过来，最后，原先的数组再被回收。可见vector容量的扩大是一个颇费时间的事。 
通常，默认的10个元素大小是不够的。你最好能准确的估计你所需要的最佳大小。 

例子： 

import java.util.vector;
public class dic {
    public void addobjects (object[] o) {
        // if length > 10, vector needs to expand
        for (int i = 0; i< o.length;i++) {
            v.add(o);   // capacity before it can add more elements.
        }
    }
    public vector v = new vector();  // no initialcapacity.
}

import java.util.vector;
public class dic {
    public void addobjects (object[] o) {
        // if length > 10, vector needs to expand
        for (int i = 0; i< o.length;i++) {    
            v.add(o);   // capacity before it can add more elements.
        }
    }
    public vector v = new vector();  // no initialcapacity.
}

更正： 
自己设定初始大小。 

public vector v = new vector(20);
public hashtable hash = new hashtable(10);

    public vector v = new vector(20);  
    public hashtable hash = new hashtable(10);

参考资料： 
dov bulka, "java performance and scalability volume 1: server-side programming 
techniques" addison wesley, isbn: 0-201-70429-3 pp.55 – 57 

三、在finally块中关闭stream 

程序中使用到的资源应当被释放，以避免资源泄漏。这最好在finally块中去做。不管程序执行的结果如何，finally块总是会执行的，以确保资源的正确关闭。 
         
例子： 

import java.io.*;
public class cs {
    public static void main (string args[]) {
        cs cs = new cs ();
        cs.method ();
    }
    public void method () {
        try {
            fileinputstream fis = new fileinputstream ("cs.java");
            int count = 0;
            while (fis.read () != -1)
                count++;
            system.out.println (count);
            fis.close ();
        } catch (filenotfoundexception e1) {
        } catch (ioexception e2) {
        }
    }
}

import java.io.*;
public class cs {
    public static void main (string args[]) {
        cs cs = new cs ();
        cs.method ();
    }
    public void method () {
        try {
            fileinputstream fis = new fileinputstream ("cs.java");
            int count = 0;
            while (fis.read () != -1)
                count++;
            system.out.println (count);
            fis.close ();
        } catch (filenotfoundexception e1) {
        } catch (ioexception e2) {
        }
    }
}

         
更正： 
在最后一个catch后添加一个finally块 

参考资料： 
peter haggar: "practical java - programming language guide". 
addison wesley, 2000, pp.77-79 
四、使用'system.arraycopy ()'代替通过来循环复制数组 

'system.arraycopy ()' 要比通过循环来复制数组快的多。 
         
例子： 

public class irb
{
    void method () {
        int[] array1 = new int [100];
        for (int i = 0; i < array1.length; i++) {
            array1 [i] = i;
        }
        int[] array2 = new int [100];
        for (int i = 0; i < array2.length; i++) {
            array2 [i] = array1 [i];                 // violation
        }
    }
}

public class irb
{
    void method () {
        int[] array1 = new int [100];
        for (int i = 0; i < array1.length; i++) {
            array1 [i] = i;
        }
        int[] array2 = new int [100];
        for (int i = 0; i < array2.length; i++) {
            array2 [i] = array1 [i];                 // violation
        }
    }
}

         
更正： 

public class irb
{
    void method () {
        int[] array1 = new int [100];
        for (int i = 0; i < array1.length; i++) {
            array1 [i] = i;
        }
        int[] array2 = new int [100];
        system.arraycopy(array1, 0, array2, 0, 100);
    }
}

public class irb
{
    void method () {
        int[] array1 = new int [100];
        for (int i = 0; i < array1.length; i++) {
            array1 [i] = i;
        }
        int[] array2 = new int [100];
        system.arraycopy(array1, 0, array2, 0, 100);
    }
}

         
参考资料： 
http://www.cs.cmu.edu/~jch/java/speed.html 

五、让访问实例内变量的getter/setter方法变成”final” 

简单的getter/setter方法应该被置成final，这会告诉编译器，这个方法不会被重载，所以，可以变成”inlined” 

例子： 

class maf {
    public void setsize (int size) {
         _size = size;
    }
    private int _size;
}

class maf {
    public void setsize (int size) {
         _size = size;
    }
    private int _size;
}

更正： 

class daf_fixed {
    final public void setsize (int size) {
         _size = size;
    }
    private int _size;
}

class daf_fixed {
    final public void setsize (int size) {
         _size = size;
    }
    private int _size;
}

参考资料： 
warren n. and bishop p. (1999), "java in practice", p. 4-5 
addison-wesley, isbn 0-201-36065-9 

六、避免不需要的instanceof操作 

如果左边的对象的静态类型等于右边的，instanceof表达式返回永远为true。 
         
例子：         

public class uiso {
    public uiso () {}
}
class dog extends uiso {
    void method (dog dog, uiso u) {
        dog d = dog;
        if (d instanceof uiso) // always true.
            system.out.println("dog is a uiso");
        uiso uiso = u;
        if (uiso instanceof object) // always true.
            system.out.println("uiso is an object");
    }
}

public class uiso {
    public uiso () {}
}
class dog extends uiso {
    void method (dog dog, uiso u) {
        dog d = dog;
        if (d instanceof uiso) // always true.
            system.out.println("dog is a uiso");
        uiso uiso = u;
        if (uiso instanceof object) // always true.
            system.out.println("uiso is an object");
    }
}

         
更正：         
删掉不需要的instanceof操作。 
         

class dog extends uiso {
    void method () {
        dog d;
        system.out.println ("dog is an uiso");
        system.out.println ("uiso is an uiso");
    }
}

class dog extends uiso {
    void method () {
        dog d;
        system.out.println ("dog is an uiso");
        system.out.println ("uiso is an uiso");
    }
}

七、避免不需要的造型操作 

所有的类都是直接或者间接继承自object。同样，所有的子类也都隐含的“等于”其父类。那么，由子类造型至父类的操作就是不必要的了。 
例子： 

class unc {
    string _id = "unc";
}
class dog extends unc {
    void method () {
        dog dog = new dog ();
        unc animal = (unc)dog;  // not necessary.
        object o = (object)dog;         // not necessary.
    }
}

class unc {
    string _id = "unc";
}
class dog extends unc {
    void method () {
        dog dog = new dog ();
        unc animal = (unc)dog;  // not necessary.
        object o = (object)dog;         // not necessary.
    }
}

         
更正：         

class dog extends unc {
    void method () {
        dog dog = new dog();
        unc animal = dog;
        object o = dog;
    }
}

class dog extends unc {
    void method () {
        dog dog = new dog();
        unc animal = dog;
        object o = dog;
    }
}
   

      
参考资料： 
nigel warren, philip bishop: "java in practice - design styles and idioms 
for effective java".  addison-wesley, 1999. pp.22-23 
八、如果只是查找单个字符的话，用charat()代替startswith() 

用一个字符作为参数调用startswith()也会工作的很好，但从性能角度上来看，调用用string api无疑是错误的! 
         
例子： 

public class pcts {
    private void method(string s) {
        if (s.startswith("a")) { // violation
            // ...
        }
    }
}

public class pcts {
    private void method(string s) {
        if (s.startswith("a")) { // violation
            // ...
        }
    }
}

         
更正         
将'startswith()' 替换成'charat()'. 

public class pcts {
    private void method(string s) {
        if ('a' == s.charat(0)) {
            // ...
        }
    }
}

public class pcts {
    private void method(string s) {
        if ('a' == s.charat(0)) {
            // ...
        }
    }
}

         
参考资料： 
dov bulka, "java performance and scalability volume 1: server-side programming 
techniques"  addison wesley, isbn: 0-201-70429-3 
九、使用移位操作来代替'a / b'操作 

"/"是一个很“昂贵”的操作，使用移位操作将会更快更有效。 

例子： 

public class sdiv {
    public static final int num = 16;
    public void calculate(int a) {
        int div = a / 4;            // should be replaced with "a >> 2".
        int div2 = a / 8;         // should be replaced with "a >> 3".
        int temp = a / 3;
    }
}

public class sdiv {
    public static final int num = 16;
    public void calculate(int a) {
        int div = a / 4;            // should be replaced with "a >> 2".
        int div2 = a / 8;         // should be replaced with "a >> 3".
        int temp = a / 3;
    }
}

更正： 

public class sdiv {
    public static final int num = 16;
    public void calculate(int a) {
        int div = a >> 2;
        int div2 = a >> 3;
        int temp = a / 3;       // 不能转换成位移操作
    }
}

public class sdiv {
    public static final int num = 16;
    public void calculate(int a) {
        int div = a >> 2;  
        int div2 = a >> 3;
        int temp = a / 3;       // 不能转换成位移操作
    }
}

十、使用移位操作代替'a * b' 

同上。 
[i]但我个人认为，除非是在一个非常大的循环内，性能非常重要，而且你很清楚你自己在做什么，方可使用这种方法。否则提高性能所带来的程序晚读性的降低将是不合算的。 

例子： 

public class smul {
    public void calculate(int a) {
        int mul = a * 4;            // should be replaced with "a << 2".
        int mul2 = 8 * a;         // should be replaced with "a << 3".
        int temp = a * 3;
    }
}

public class smul {
    public void calculate(int a) {
        int mul = a * 4;            // should be replaced with "a << 2".
        int mul2 = 8 * a;         // should be replaced with "a << 3".
        int temp = a * 3;
    }
}

更正： 

package opt;
public class smul {
    public void calculate(int a) {
        int mul = a << 2;
        int mul2 = a << 3;
        int temp = a * 3;       // 不能转换
    }
}

package opt;
public class smul {
    public void calculate(int a) {
        int mul = a << 2;  
        int mul2 = a << 3;
        int temp = a * 3;       // 不能转换
    }
}

十一、在字符串相加的时候，使用 ' ' 代替 " "，如果该字符串只有一个字符的话 


例子： 

public class str {
    public void method(string s) {
        string string = s + "d"  // violation.
        string = "abc" + "d"      // violation.
    }
}

public class str {
    public void method(string s) {
        string string = s + "d"  // violation.
        string = "abc" + "d"      // violation.
    }
}

更正： 
将一个字符的字符串替换成' ' 

public class str {
    public void method(string s) {
        string string = s + 'd'
        string = "abc" + 'd'
    }
}

public class str {
    public void method(string s) {
        string string = s + 'd'
        string = "abc" + 'd'   
    }
}

十二、不要在循环中调用synchronized(同步)方法 

方法的同步需要消耗相当大的资料，在一个循环中调用它绝对不是一个好主意。 

例子： 

import java.util.vector;
public class syn {
    public synchronized void method (object o) {
    }
    private void test () {
        for (int i = 0; i < vector.size(); i++) {
            method (vector.elementat(i));    // violation
        }
    }
    private vector vector = new vector (5, 5);
}

import java.util.vector;
public class syn {
    public synchronized void method (object o) {
    }
    private void test () {
        for (int i = 0; i < vector.size(); i++) {
            method (vector.elementat(i));    // violation
        }
    }
    private vector vector = new vector (5, 5);
}

更正： 
不要在循环体中调用同步方法，如果必须同步的话，推荐以下方式： 

import java.util.vector;
public class syn {
    public void method (object o) {
    }
private void test () {
    synchronized{//在一个同步块中执行非同步方法
            for (int i = 0; i < vector.size(); i++) {
                method (vector.elementat(i));
            }
        }
    }
    private vector vector = new vector (5, 5);
}

import java.util.vector;
public class syn {
    public void method (object o) {
    }
private void test () {
    synchronized{//在一个同步块中执行非同步方法
            for (int i = 0; i < vector.size(); i++) {
                method (vector.elementat(i));   
            }
        }
    }
    private vector vector = new vector (5, 5);
}

十三、将try/catch块移出循环 

把try/catch块放入循环体内，会极大的影响性能，如果编译jit被关闭或者你所使用的是一个不带jit的jvm，性能会将下降21%之多! 
         
例子：         

import java.io.fileinputstream;
public class try {
    void method (fileinputstream fis) {
        for (int i = 0; i < size; i++) {
            try {                                      // violation
                _sum += fis.read();
            } catch (exception e) {}
        }
    }
    private int _sum;
}

import java.io.fileinputstream;
public class try {
    void method (fileinputstream fis) {
        for (int i = 0; i < size; i++) {
            try {                                      // violation
                _sum += fis.read();
            } catch (exception e) {}
        }
    }
    private int _sum;
}

         
更正：         
将try/catch块移出循环         
  

void method (fileinputstream fis) {
       try {
           for (int i = 0; i < size; i++) {
               _sum += fis.read();
           }
       } catch (exception e) {}
   }

 void method (fileinputstream fis) {
        try {
            for (int i = 0; i < size; i++) {
                _sum += fis.read();
            }
        } catch (exception e) {}
    }

         
参考资料： 
peter haggar: "practical java - programming language guide". 
addison wesley, 2000, pp.81 – 83 

十四、对于boolean值，避免不必要的等式判断 

将一个boolean值与一个true比较是一个恒等操作(直接返回该boolean变量的值). 移走对于boolean的不必要操作至少会带来2个好处： 
1)代码执行的更快 (生成的字节码少了5个字节)； 
2)代码也会更加干净 。 

例子： 

public class ueq
{
    boolean method (string string) {
        return string.endswith ("a") == true;   // violation
    }
}

public class ueq
{
    boolean method (string string) {
        return string.endswith ("a") == true;   // violation
    }
}

更正： 

class ueq_fixed
{
    boolean method (string string) {
        return string.endswith ("a");
    }
}

class ueq_fixed
{
    boolean method (string string) {
        return string.endswith ("a");
    }
}

十五、对于常量字符串，用'string' 代替 'stringbuffer' 

常量字符串并不需要动态改变长度。 
例子： 

public class usc {
    string method () {
        stringbuffer s = new stringbuffer ("hello");
        string t = s + "world!";
        return t;
    }
}

public class usc {
    string method () {
        stringbuffer s = new stringbuffer ("hello");
        string t = s + "world!";
        return t;
    }
}

更正： 
把stringbuffer换成string，如果确定这个string不会再变的话，这将会减少运行开销提高性能。 

十六、用'stringtokenizer' 代替 'indexof()' 和'substring()' 

字符串的分析在很多应用中都是常见的。使用indexof()和substring()来分析字符串容易导致 stringindexoutofboundsexception。而使用stringtokenizer类来分析字符串则会容易一些，效率也会高一些。 

例子： 

public class ust {
    void parsestring(string string) {
        int index = 0;
        while ((index = string.indexof(".", index)) != -1) {
            system.out.println (string.substring(index, string.length()));
        }
    }
}

public class ust {
    void parsestring(string string) {
        int index = 0;
        while ((index = string.indexof(".", index)) != -1) {
            system.out.println (string.substring(index, string.length()));
        }
    }
}

参考资料： 
graig larman, rhett guthrie: "java 2 performance and idiom guide" 
prentice hall ptr, isbn: 0-13-014260-3 pp. 282 – 283 

十七、使用条件操作符替代"if (cond) return; else return;" 结构 

条件操作符更加的简捷 
例子： 

public class if {
    public int method(boolean isdone) {
        if (isdone) {
            return 0;
        } else {
            return 10;
        }
    }
}

public class if {
    public int method(boolean isdone) {
        if (isdone) {
            return 0;
        } else {
            return 10;
        }
    }
}

更正： 

public class if {
    public int method(boolean isdone) {
        return (isdone ? 0 : 10);
    }
}

public class if {
    public int method(boolean isdone) {
        return (isdone ? 0 : 10);
    }
}

十八、使用条件操作符代替"if (cond) a = b; else a = c;" 结构 

例子： 

public class ifas {
    void method(boolean istrue) {
        if (istrue) {
            _value = 0;
        } else {
            _value = 1;
        }
    }
    private int _value = 0;
}

public class ifas {
    void method(boolean istrue) {
        if (istrue) {
            _value = 0;
        } else {
            _value = 1;
        }
    }
    private int _value = 0;
}

更正： 

public class ifas {
    void method(boolean istrue) {
        _value = (istrue ? 0 : 1);       // compact expression.
    }
    private int _value = 0;
}

public class ifas {
    void method(boolean istrue) {
        _value = (istrue ? 0 : 1);       // compact expression.
    }
    private int _value = 0;
}

十九、不要在循环体中实例化变量 

在循环体中实例化临时变量将会增加内存消耗 

例子：         

import java.util.vector;
public class loop {
    void method (vector v) {
        for (int i=0;i < v.size();i++) {
            object o = new object();
            o = v.elementat(i);
        }
    }
}

import java.util.vector;
public class loop {
    void method (vector v) {
        for (int i=0;i < v.size();i++) {
            object o = new object();
            o = v.elementat(i);
        }
    }
}

         
更正：         
在循环体外定义变量，并反复使用         

import java.util.vector;
public class loop {
    void method (vector v) {
        object o;
        for (int i=0;i<v.size();i++) {
            o = v.elementat(i);
        }
    }
}

import java.util.vector;
public class loop {
    void method (vector v) {
        object o;
        for (int i=0;i<v.size();i++) {
            o = v.elementat(i);
        }
    }
}

二十、确定 stringbuffer的容量 

stringbuffer的构造器会创建一个默认大小(通常是16)的字符数组。在使用中，如果超出这个大小，就会重新分配内存，创建一个更大的数组，并将原先的数组复制过来，再丢弃旧的数组。在大多数情况下，你可以在创建stringbuffer的时候指定大小，这样就避免了在容量不够的时候自动增长，以提高性能。 

例子：         

public class rsbc {
    void method () {
        stringbuffer buffer = new stringbuffer(); // violation
        buffer.append ("hello");
    }
}

public class rsbc {
    void method () {
        stringbuffer buffer = new stringbuffer(); // violation
        buffer.append ("hello");
    }
}

         
更正：         
为stringbuffer提供寝大小。         

public class rsbc {
    void method () {
        stringbuffer buffer = new stringbuffer(max);
        buffer.append ("hello");
    }
    private final int max = 100;
}

public class rsbc {
    void method () {
        stringbuffer buffer = new stringbuffer(max);
        buffer.append ("hello");
    }
    private final int max = 100;
}

         
参考资料： 
dov bulka, "java performance and scalability volume 1: server-side programming 
techniques" addison wesley, isbn: 0-201-70429-3 p.30 – 31 

二十一、尽可能的使用栈变量 

如果一个变量需要经常访问，那么你就需要考虑这个变量的作用域了。static? local?还是实例变量？访问静态变量和实例变量将会比访问局部变量多耗费2-3个时钟周期。 
         
例子： 

public class usv {
    void getsum (int[] values) {
        for (int i=0; i < value.length; i++) {
            _sum += value[i];           // violation.
        }
    }
    void getsum2 (int[] values) {
        for (int i=0; i < value.length; i++) {
            _staticsum += value[i];
        }
    }
    private int _sum;
    private static int _staticsum;
}

public class usv {
    void getsum (int[] values) {
        for (int i=0; i < value.length; i++) {
            _sum += value[i];           // violation.
        }
    }
    void getsum2 (int[] values) {
        for (int i=0; i < value.length; i++) {
            _staticsum += value[i];
        }
    }
    private int _sum;
    private static int _staticsum;
}     

         
更正：         
如果可能，请使用局部变量作为你经常访问的变量。 
你可以按下面的方法来修改getsum()方法：         

void getsum (int[] values) {
    int sum = _sum;  // temporary local variable.
    for (int i=0; i < value.length; i++) {
        sum += value[i];
    }
    _sum = sum;
}

void getsum (int[] values) {
    int sum = _sum;  // temporary local variable.
    for (int i=0; i < value.length; i++) {
        sum += value[i];
    }
    _sum = sum;
}

         
参考资料：         
peter haggar: "practical java - programming language guide". 
addison wesley, 2000, pp.122 – 125 

二十二、不要总是使用取反操作符(!) 

取反操作符(!)降低程序的可读性，所以不要总是使用。 

例子： 

public class dun {
    boolean method (boolean a, boolean b) {
        if (!a)
            return !a;
        else
            return !b;
    }
}

public class dun {
    boolean method (boolean a, boolean b) {
        if (!a)
            return !a;
        else
            return !b;
    }
}

更正： 
如果可能不要使用取反操作符(!) 

二十三、与一个接口 进行instanceof操作 

基于接口的设计通常是件好事，因为它允许有不同的实现，而又保持灵活。只要可能，对一个对象进行instanceof操作，以判断它是否某一接口要比是否某一个类要快。 

例子： 

public class insof {
    private void method (object o) {
        if (o instanceof interfacebase) { }  // better
        if (o instanceof classbase) { }   // worse.
    }
}

class classbase {}
interface interfacebase {}

public class insof {
    private void method (object o) {
        if (o instanceof interfacebase) { }  // better
        if (o instanceof classbase) { }   // worse.
    }
}

class classbase {}
interface interfacebase {}

参考资料： 
graig larman, rhett guthrie: "java 2 performance and idiom guide" 
prentice hall ptr, 2000.  pp.207