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: queens.groovy int q=9 int[] i=new int[q] int count=0 long t = System.currentTimeMillis(); scan(0) println("totle results:"+count) println("totle time:"+(System.currentTimeMillis()-t)); def scan(n){ if (n==q){ println(i.toList()) count++ return } i[n]=0 while(i[n]<q){ i[n] = i[n]+1 if (check(n)) scan(n+1) } } def check(n){ if (n>0) for (j in 0..n-1) if (i[j]==i[n] || i[j]-i[n]==j-n || i[j]-i[n]==n-j ) return false return true } 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: public class queens { static int q=9; static int[] i=new int[q]; static int count=0; public static void main(String[] args){ long t = System.currentTimeMillis(); scan(0); System.out.println("totle results:"+count); System.out.println("totle time:"+(System.currentTimeMillis()-t)); } private static void scan(int n){ if (n==q){ for (int k=0;k<q;k++) System.out.print(i[k]+(k==q-1?"\n":",")); count++; return; } i[n]=0; while(i[n]<q){ i[n] = i[n]+1; if (check(n)){ scan(n+1); } } } private static boolean check(int n){ int j=0; while(j<n){ if (i[j]==i[n] || i[j]-i[n]==j-n || i[j]-i[n]==n-j ){ return false; } j++; } return true; } } 运行结果是:totle time:791 每次运行花费的时间略有不同,大致相差10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: <SCRIPT LANGUAGE="JavaScript"> <!-- var q=9 var i=[] var count=0 var d = new Date(); scan(0) document.write("totle results:"+count+"<br>") document.write("time used:"+(new Date()-d)+"<br>") function scan(n){ if (n==q){ document.write(i+"<br>") count++ return } i[n]=0 while(i[n]<q){ i[n] = i[n]+1 if (check(n)){ scan(n+1) } } } function check(n){ var j=0; while(j<n){ if (i[j]==i[n] || i[j]-i[n]==j-n || i[j]-i[n]==n-j ){ return false } j++ } return true } //--> </SCRIPT> 运行结果是: time used:1241 只比java慢了不到一倍,看来groovy还差的太远了。 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函数把: def check(n){ if (n>0) for (j in 0..n-1) if (i[j]==i[n] || i[j]-i[n]==j-n || i[j]-i[n]==n-j ) return false return true } 编译后变成 public Object check(Object obj) { if(ScriptBytecodeAdapter.compareGreaterThan(obj, new Integer(0))) { Object obj1 = null; for(Iterator iterator = ScriptBytecodeAdapter.asIterator(ScriptBytecodeAdapter.createRange(new Integer(0), ScriptBytecodeAdapter.invokeMethod(obj, "minus", ((Object) (new Object[] { new Integer(1) }))), true)); iterator.hasNext();) { Object obj2 = iterator.next(); Object obj3 = null; if(ScriptBytecodeAdapter.asBool(ScriptBytecodeAdapter.asBool(ScriptBytecodeAdapter.compareEqual(ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] { obj2 }))), ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] { obj })))) || ScriptBytecodeAdapter.compareEqual(ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] { obj2 }))), "minus", ((Object) (new Object[] { ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] { obj }))) }))), ScriptBytecodeAdapter.invokeMethod(obj2, "minus", ((Object) (new Object[] { obj })))) ? ((Object) (Boolean.TRUE)) : ((Object) (Boolean.FALSE))) || ScriptBytecodeAdapter.compareEqual(ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] { obj2 }))), "minus", ((Object) (new Object[] { ScriptBytecodeAdapter.invokeMethod(ScriptBytecodeAdapter.getGroovyObjectProperty(this, "i"), "getAt", ((Object) (new Object[] { obj }))) }))), ScriptBytecodeAdapter.invokeMethod(obj, "minus", ((Object) (new Object[] { obj2 })))) ? ((Object) (Boolean.TRUE)) : ((Object) (Boolean.FALSE)))) return Boolean.FALSE; } } return Boolean.TRUE; } 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? 为了减少编码量要花费10倍的运行效率做代价,我认为这个代价太高了。还是等待groovy的继续改进吧。

|