第一篇笔记里面,我说groovy运行的居然还满快的,其实是个误会了。我上次做八皇后还是在8080上面用basic做的,和现在奔四上面的groovy相比是没有意义的。特地又做了个对比试验:
1
int q=9
2
int[] i=new int[q]
3
int count=0
4
long t = System.currentTimeMillis();
5
scan(0)
6
println("totle results:"+count)
7
println("totle time:"+(System.currentTimeMillis()-t));
8
def scan(n){
9
if (n==q){
10
println(i.toList())
11
count++
12
return
13
}
14
i[n]=0
15
while(i[n]<q){
16
i[n] = i[n]+1
17
if (check(n))
18
scan(n+1)
19
}
20
}
21
def check(n){
22
if (n>0)
23
for (j in 0..<n)
24
if (i[j]==i[n] || i[j]-i[n]==j-n || i[j]-i[n]==n-j )
25
return false
26
return true
27
}

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

运行结果是:totle time:7271 (为了用groovy控制台运行的,直接用groovy命令运行还要慢一点)
java呢?
queens.java:
1
public class queens {
2
static int q=9;
3
static int[] i=new int[q];
4
static int count=0;
5
public static void main(String[] args){
6
long t = System.currentTimeMillis();
7
scan(0);
8
System.out.println("totle results:"+count);
9
System.out.println("totle time:"+(System.currentTimeMillis()-t));
10
}
11
private static void scan(int n){
12
if (n==q){
13
for (int k=0;k<q;k++) System.out.print(i[k]+(k==q-1?"/n":","));
14
count++;
15
return;
16
}
17
i[n]=0;
18
while(i[n]<q){
19
i[n] = i[n]+1;
20
if (check(n)){
21
scan(n+1);
22
}
23
}
24
}
25
private static boolean check(int n){
26
for(int j=0;j<n;j++){
27
if (i[j]==i[n] || i[j]-i[n]==j-n || i[j]-i[n]==n-j ){
28
return false;
29
}
30
}
31
return true;
32
}
33
}
34

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

运行结果是:totle time:271
每次运行花费的时间略有不同,groovy和java的运行速度看来大致相差10~30倍左右。
能说这是脚本语言天生的缺陷吗?我们来看看同样是类似java语法的脚本语言javascript在IE里面的速度:
1
var q=9
2
var i=[]
3
var count=0
4
var d = new Date();
5
scan(0)
6
document.write("totle results:"+count+"<br>")
7
document.write("time used:"+(new Date()-d)+"<br>")
8
9
function scan(n){
10
if (n==q){
11
document.write(i+"<br>")
12
count++
13
return
14
}
15
i[n]=0
16
while(i[n]<q){
17
i[n] = i[n]+1
18
if (check(n)){
19
scan(n+1)
20
}
21
}
22
}
23
24
function check(n){
25
for (var j=0; j<n;j++)
26
if (i[j]==i[n] || i[j]-i[n]==j-n || i[j]-i[n]==n-j )
27
return false
28
return true
29
}

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

运行结果是: time used:1241
比groovy快了5倍以上。groovy可真是够慢的。
把groovy编译的class文件反编译了一下,看到groovy生成的代码效率确实是太低了,我们就看循环最内层的check函数吧:
1
def check(n){
2
if (n>0)
3
for (j in 0..<n)
4
if (i[j]==i[n] || i[j]-i[n]==j-n || i[j]-i[n]==n-j )
5
return false
6
return true
7
}

2

3

4

5

6

7

编译后变成
1
public Object check(Object obj)
2
{
3
if(ScriptBytecodeAdapter.compareGreaterThan(obj, new Integer(0)))
4
{
5
Object obj1 = null;
6
for(Iterator iterator = ScriptBytecodeAdapter.asIterator(ScriptBytecodeAdapter.createRange(new Integer(0), obj, false)); iterator.hasNext();)
7
{
8
Object obj2 = iterator.next();
9
Object obj3 = null;
10
if(ScriptBytecodeAdapter.asBool(ScriptBytecodeAdapter.asBool(ScriptBytecodeAdapter.compareEqual(ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] {
11
obj2
12
}))), ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] {
13
obj
14
})))) || ScriptBytecodeAdapter.compareEqual(ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] {
15
obj2
16
}))), "minus", ((Object) (new Object[] {
17
ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] {
18
obj
19
})))
20
}))), ScriptBytecodeAdapter.invokeMethod(obj2, "minus", ((Object) (new Object[] {
21
obj
22
})))) ? ((Object) (Boolean.TRUE)) : ((Object) (Boolean.FALSE))) || ScriptBytecodeAdapter.compareEqual(ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] {
23
obj2
24
}))), "minus", ((Object) (new Object[] {
25
ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] {
26
obj
27
})))
28
}))), ScriptBytecodeAdapter.invokeMethod(obj, "minus", ((Object) (new Object[] {
29
obj2
30
})))) ? ((Object) (Boolean.TRUE)) : ((Object) (Boolean.FALSE))))
31
return Boolean.FALSE;
32
}
33
34
}
35
return Boolean.TRUE;
36
}
37

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

一切都是object,做任何事情都是invokeMethod,两个整数的比较居然要写将近400个字符的代码,光看代码量都可以吓倒我了。这是我们期待的脚本语言吗?
groovy可以嵌入到java代码里面,但是java代码可以嵌入到groovy里面吗?我觉得groovy有必要提供这样一种机制,在有必要的时候可以消除性能瓶颈。可是现在只看到groovy里面可以通过Scriptom(现在还是beta版)嵌入vbs、js脚本(包括使用WSH,FSO)或者调用InternetExplorer、Media Player、Word和Excel等windows组件。看来对消除性能瓶颈的帮助不大。