Class文件详解【转】

http://blog.youkuaiyun.com/zhy05/archive/2007/09/24/1798843.aspx

 

 

我们都知道，Java编译器负责将.java文件编译成.class文件，class文件存储的是java字节码，与.java文件无关（只要你愿意写一个编译器，也可以将别的语言写的源代码编译成.class文件），本文准备详细解剖class文件的内部结构，并且把class文件结构读取并显示出来。

Class文件的格式由JVM规范规定，一共有以下部分：
1. magic number，必须是0xCAFEBABE，用于快速识别是否是一个class文件。
2. version，包括major和minor，如果版本号超过了JVM的识别范围，JVM将拒绝执行。
3. constant pool，常量池，存放所有用到的常量。
4. access flag，定义类的访问权限。
5. this class和super class，指示如何找到this class和super class。
6. interfaces，存放所有interfaces。
7. fields，存放所有fields。
8. methods，存放所有methods。
9. attributes，存放所有attributes。

先写一个Test.java：

package example.test;

public final class TestClass {
    public int id = 1234567;
    public void test() {}
}

然后编译，放在C:\example\test\Test.class。

我们用Java来读取和分析class，ClassAnalyzer的功能便是读取Test.class，分析结构，然后显示出来：

package classfile.format;

import java.io.*;

public class ClassAnalyzer {

    public static void main(String[] args) {
        DataInputStream input = null;
        try {
            input = new DataInputStream(new BufferedInputStream(new FileInputStream(
                "C:\\example\\test\\TestClass.class"
            )));
            analyze(input);
        }
        catch(Exception e) {
            System.out.println("Analyze failed!");
        }
        finally {
            try { input.close(); } catch(Exception e) {}
        }
    }

    public static void analyze(DataInputStream input) throws IOException {
        // read magic number:
        int magic = input.readInt();
        if(magic==0xCAFEBABE)
            System.out.println("magic number = 0xCAFEBABE");
        else
            throw new RuntimeException("Invalid magic number!");
        // read minor version and major version:
        short minor_ver = input.readShort();
        short major_ver = input.readShort();
        System.out.println("Version = " + major_ver + "." + minor_ver);
        // read constant pool:
        short const_pool_count = input.readShort();
        System.out.println("constant pool size = " + const_pool_count);
        // read each constant:
        for(int i=1; i<const_pool_count; i++) {
            analyzeConstant(input, i);
        }
    }

    public static void analyzeConstant(DataInputStream input, int index) throws IOException {
        byte flag = input.readByte();
        // for read:
        byte n8;
        short n16;
        int n32;
        long n64;
        float f;
        double d;
        byte[] buffer;
        System.out.println("\nconst index = " + index + ", flag = " + (int)flag);
        switch(flag) {
        case 1: // utf-8 string
            System.out.println(" const type = Utf8");
            n16 = input.readShort();
            System.out.println("     length = " + n16);
            buffer = new byte[n16];
            input.readFully(buffer);
            System.out.println("      value = " + new String(buffer));
            break;
        case 3: // integer
            System.out.println(" const type = Integer");
            n32 = input.readInt();
            System.out.println("      value = " + n32);
            break;
        case 4: // float
            System.out.println(" const type = Float");
            f = input.readFloat();
            System.out.println("      value = " + f);
            break;
        case 5: // long
            System.out.println(" const type = Long");
            n64 = input.readLong();
            System.out.println("      value = " + n64);
            break;
        case 6: // double
            System.out.println(" const type = Double");
            d = input.readDouble();
            System.out.println("      value = " + d);
            break;
        case 7: // class or interface reference
            System.out.println(" const type = Class");
            n16 = input.readShort();
            System.out.println("      index = " + n16 + " (where to find the class name)");
            break;
        case 8: // string
            System.out.println(" const type = String");
            n16 = input.readShort();
            System.out.println("      index = " + n16);
            break;
        case 9: // field reference
            System.out.println(" const type = Fieldref");
            n16 = input.readShort();
            System.out.println("class index = " + n16 + " (where to find the class)");
            n16 = input.readShort();
            System.out.println("nameAndType = " + n16 + " (where to find the NameAndType)");
            break;
        case 10: // method reference
            System.out.println(" const type = Methodref");
            n16 = input.readShort();
            System.out.println("class index = " + n16 + " (where to find the class)");
            n16 = input.readShort();
            System.out.println("nameAndType = " + n16 + " (where to find the NameAndType)");
            break;
        case 11: // interface method reference
            System.out.println(" const type = InterfaceMethodref");
            n16 = input.readShort();
            System.out.println("class index = " + n16 + " (where to find the interface)");
            n16 = input.readShort();
            System.out.println("nameAndType = " + n16 + " (where to find the NameAndType)");
            break;
        case 12: // name and type reference
            System.out.println(" const type = NameAndType");
            n16 = input.readShort();
            System.out.println(" name index = " + n16 + " (where to find the name)");
            n16 = input.readShort();
            System.out.println(" descripter = " + n16 + " (where to find the descriptor)");
            break;
        default:
            throw new RuntimeException("Invalid constant pool flag: " + flag);
        }
    }
}

输出结果为：

magic number = 0xCAFEBABE
Version = 48.0
constant pool size = 22

const index = 1, flag = 1
 const type = Utf8
     length = 22
      value = example/test/TestClass

const index = 2, flag = 7
 const type = Class
      index = 1 (where to find the class name)

const index = 3, flag = 1
 const type = Utf8
     length = 16
      value = java/lang/Object

const index = 4, flag = 7
 const type = Class
      index = 3 (where to find the class name)

const index = 5, flag = 1
 const type = Utf8
     length = 2
      value = id

const index = 6, flag = 1
 const type = Utf8
     length = 1
      value = I

const index = 7, flag = 1
 const type = Utf8
     length = 6
      value = <init>

const index = 8, flag = 1
 const type = Utf8
     length = 3
      value = ()V

const index = 9, flag = 1
 const type = Utf8
     length = 4
      value = Code

const index = 10, flag = 12
 const type = NameAndType
 name index = 7 (where to find the name)
 descripter = 8 (where to find the descriptor)

const index = 11, flag = 10
 const type = Methodref
class index = 4 (where to find the class)
nameAndType = 10 (where to find the NameAndType)

const index = 12, flag = 3
 const type = Integer
      value = 1234567

const index = 13, flag = 12
 const type = NameAndType
 name index = 5 (where to find the name)
 descripter = 6 (where to find the descriptor)

const index = 14, flag = 9
 const type = Fieldref
class index = 2 (where to find the class)
nameAndType = 13 (where to find the NameAndType)

const index = 15, flag = 1
 const type = Utf8
     length = 15
      value = LineNumberTable

const index = 16, flag = 1
 const type = Utf8
     length = 18
      value = LocalVariableTable

const index = 17, flag = 1
 const type = Utf8
     length = 4
      value = this

const index = 18, flag = 1
 const type = Utf8
     length = 24
      value = Lexample/test/TestClass;

const index = 19, flag = 1
 const type = Utf8
     length = 4
      value = test

const index = 20, flag = 1
 const type = Utf8
     length = 10
      value = SourceFile

const index = 21, flag = 1
 const type = Utf8
     length = 14
      value = TestClass.java

我们暂时只读取到constant pool，后面的信息下次再继续 :)

 

 

 

 

继续分析Class文件的结构。
　　
　　上次读取了constant pool，紧接着是这个类或接口的Access flags，JVM定义的关于class或interface的Access flags有：
　　
　　private static short ACC_PUBLIC　　= 0x0001;
　　private static short ACC_FINAL　　 = 0x0010;
　　private static short ACC_SUPER　　 = 0x0020;
　　private static short ACC_INTERFACE = 0x0200;
　　private static short ACC_ABSTRACT　= 0x0400;
　　
　　// read access flags:
　　short access_flags = input.readShort();
　　System.out.print("Access flags:");
　　if((access_flags & ACC_PUBLIC) == ACC_PUBLIC)
　　System.out.print(" public");
　　if((access_flags & ACC_FINAL) == ACC_FINAL)
　　System.out.print(" final");
　　if((access_flags & ACC_SUPER) == ACC_SUPER)
　　System.out.print(" super");
　　if((access_flags & ACC_INTERFACE) == ACC_INTERFACE)
　　System.out.print(" interface");
　　if((access_flags & ACC_ABSTRACT) == ACC_ABSTRACT)
　　System.out.print(" abstract");
　　System.out.println(); feedom.net
　　
　　然后是this class和super class：
　　
　　// read this class and super class:
　　short this_class_index = input.readShort();
　　short super_class_index = input.readShort();
　　System.out.println("This class = " + this_class_index);
　　System.out.println("Super class = " + super_class_index);
　　
　　根据this class的index可以从constant pool中得到这个class的信息，super class也一样。每个class都有super class，只有Object的super class index=0。
　　
　　接下来是这个class继承了多少个interface和每个interface在constant pool中的index：
　　
　　// read interfaces count:
　　short interfaces_count = input.readShort();
　　System.out.println("Interfaces count = " + interfaces_count);
　　// read each interface:
　　for(int i=1; i<=interfaces_count; i++) {
　　short interface_index = input.readShort();
　　System.out.println("No. " + i + " interface index = " + interface_index);
　　}
　　
　　结果如下：
　　
　　Access flags: public final super 中国网管论坛bbs.bitsCN.com
　　This class = 2
　　Super class = 4
　　Interfaces count = 0