Beetl1.2M1将模板的文本先以Byte方式输出，性能又大幅提高（更新）

还是以以前的模板为准（大小6K），循环渲染50000次，需要时间（毫秒为单位）如下

Beetl1.2M1 解释执行方式（普通模式），运行三次，分别 1356,1365,1348
Beetl1.2M1 编译执行方式（普通模式），运行三次，分别 913,922,905
Freemarker 分别是 1155,1130,1122

Beetl1.2M1 允许字节流优化，解释方式执行三次，分别是 587,605,610

Beetl1.2M1 允许字节流优化，编译方式执行三次，分别是 385,355,370

如下式一个采用节流优化，编译方式执行 的class代码

行17 输出V0，此处借鉴了Freemarker合并多行，一次性输出的功能，此优化对性能影响也很大

行18-32 是对includeFileTemplate 标签的处理，类似JSP的include

行39-42 是将模板中的文本预先转化成字节数组

行48 是行映射，也就是在渲染出错的时候，可以将java报错的行数与模板文件行数对应起来，以方便的查看错误原因，如19=3 表示java代码19行如果抛错，则对应的是模板中的第三行错误

package ext;
import java.util.*;
import java.util.Map.Entry;
import java.math.*;
import java.io.*;
import org.bee.tl.core.*;
import org.bee.tl.core.io.*;
import org.bee.tl.core.compile.*;
/* /ext/master_template.html is generated by beetl on 2012-06-22 10:42:38 */
public class master_template_html extends CompiledClass{
    public void service(ByteWriter writer,Context ctx) throws IOException,BeeException{
        ctx.set("__group",group);
        ctx.set("__this",template);
        CompileTemplateWriter out = new CompileTemplateWriter(writer,this);
        ctx.set("__pw",out);
        try{
            out.write(__V0);
            ByteSupportTag includeFileTemplate_18_19 = (ByteSupportTag)this.getTag("includeFileTemplate");
            includeFileTemplate_18_19.setParas(new Object[]{"/ext/child_template.html"});
            includeFileTemplate_18_19.setContext(ctx.getNewContext());
            if(includeFileTemplate_18_19.requriedInput()){
                out= out.getTempWriter();
                ctx.set("__pw",out);
                {
                    out.write(__V1);
                }
                //还原输出流
                includeFileTemplate_18_19.setByteInput(out.getTempByte());
                out = out.getParent();
                ctx.set("__pw",out);
            }
            out.write(includeFileTemplate_18_19.getOutputAsByte());
            out.write(__V2);
        }catch(Exception ex){
            throw getException(ex,lineMap);
        }
        out.flush();
    }
    private static final byte[] __V1 = new byte[]{0x68,0x65,0x6c,0x6c,0x6f,0x2c,0x74,0x68,0x69,0x73,0x20,0x69,0x73,0x20,0x74,0x65,0x6d,0x70,0x20,0x63,0x68,0x69,0x6c,0x64,0xd,0xa};
    private static final byte[] __V0 = new byte[]{0xffffffd5,0xffffffe2,0xffffffca,0xffffffc7,0x6d,0x61,0x73,0x74,0x65,0x72,0xffffffd2,0xffffffb3,0xffffffc3,0xffffffe6,0xd,0xa,0xffffffcf,0xffffffd6,0xffffffd4,0xffffffda,0xffffffcf,0xffffffd4,0xffffffca,0xffffffbe,0x63,0x68,0x69,0x6c,0x64,0xffffffd2,0xffffffb3,0xffffffc3,0xffffffe6,0xffffffd4,0xffffffda,0xffffffcf,0xffffffc2,0xffffffc3,0xffffffe6,0xd,0xa};
    private static final byte[] __V2 = new byte[]{0x3d,0x3d,0x3d,0x3d,0x3d,0x3d,0x3d,0x3d,0x3d,0x3d,0x3d,0x3d,0x3d,0x3d,0x3d,0xd,0xa,0xffffffcf,0xffffffd6,0xffffffd4,0xffffffda,0xffffffca,0xffffffc7,0x6d,0x61,0x73,0x74,0x65,0x72,0xffffffd2,0xffffffb3,0xffffffc3,0xffffffe6};
    private static final byte[] __VCR = new byte[]{0xd,0xa};
    /* 原模板Cached标示*/
    public long getVersion(){
        return 1337956717968l;
    }
    /* 行映射*/
    protected String lineMap = "-32=3-19=3-21=3-";
}

可见，字节流优化对性能影响很大，最快方式是beetl编译方式，且允许字节流优化。超过了Freemarker4倍。

对于使用者来说，只需俩行命令就可以了

group.enableOptimize(); //编译成class运行

group.enableDirectOutputByte(); //使用字节流输出