前言

我在[003]AIDL是什么中介绍的AIDL，但是好像还有朋友不明白问我，那我就来写一个终极版的文章，让你十分钟彻底明白AIDL，以下代码全为手写。

目标

Server进程注册一个Calculator服务到ServiceManager，该Calculator服务提供add和minus两个接口
Client进程通过ServiceManager获得Calculator服务的代理类，通过Binder调用add和minus两个接口

1、没有AIDL的世界

1.1 Server进程

    class Calculator extends Binder {

        private static final int CODE_ADD = 1;

        private static final int CODE_MINUS = 2;

        @Override
        protected boolean onTransact(int code, @NonNull Parcel data, @Nullable Parcel reply, int flags) throws RemoteException {
            if (code == CODE_ADD) {
                int a = data.readInt();
                int b = data.readInt();
                reply.writeInt(a + b);
                return true;
            } else if (code == CODE_MINUS) {
                int a = data.readInt();
                int b = data.readInt();
                reply.writeInt(a + b);
                return true;
            }
            return super.onTransact(code, data, reply, flags);
        }
    }

    void main() {
        //注册binder服务
        ServiceManager.addService("calculator", new Calculator());
    }

1.2 Client进程

    private static final int CODE_ADD = 1;
    private static final int CODE_MINUS = 2;

    int add(int a, int b) {
        IBinder mClient = ServiceManager.getService("calculator");//BinderProxy
        Parcel data = Parcel.obtain();
        Parcel reply = Parcel.obtain();
        data.writeInt(a);
        data.writeInt(b);
        mClient.transact(CODE_ADD, data, reply, 0);
        int result = reply.readInt();
        return result;
    }

    int minus(int a, int b) {
        IBinder mClient = ServiceManager.getService("calculator");//BinderProxy
        Parcel data = Parcel.obtain();
        Parcel reply = Parcel.obtain();
        data.writeInt(a);
        data.writeInt(b);
        mClient.transact(CODE_MINUS, data, reply, 0);
        int result = reply.readInt();
        return result;
    }

2、有AIDL的世界

ICalculator.aidl文件

    //aidl
    interface ICalculator {
        int add(int a, int b);
        int minus(int a, int b);
    }

2.1 Server进程

与1.1中代码比较，我们发现，我们不是直接继承Binder，而是继承ICalculator.Stub，我们只需专注写add和minus接口的实现，不需要在Binder类中的onTransact写一堆if else和Parcel.read write

    class Calculator extends ICalculator.Stub {
        @Override
        public int add(int a, int b) {
            return a + b;
        }

        @Override
        public int minus(int a, int b) {
            return a - b;
        }
    }

    void main() {
        //注册binder服务
        ServiceManager.addService("calculator", new Calculator());
    }

2.2 Client进程

与1.2中代码比较，我们发现，我们不需要写Parcel.read write和BinderProxy.transact，而是用BinderProxy对象构造一个ICalculator.Stub.Proxy对象，然后直接调用ICalculator.Stub.Proxy的add和minus

   int add(int a, int b) {
        IBinder mClient = ServiceManager.getService("calculator");//BinderProxy
        return new ICalculator.Stub.Proxy(mClient).add(a, b);
    }

    int minus(int a, int b) {
        IBinder mClient = ServiceManager.getService("calculator");//BinderProxy
        return new ICalculator.Stub.Proxy(mClient).minus(a, b);
    }

3、AIDL自动生成了什么代码

ICalculator.aidl生成的代码如下，有三个部分组成，我们后面仔细分析

ICalculator接口
ICalculator.Stub类
ICalculator.Stub.Proxy类

interface ICalculator {//看3.1分析

        int add(int a, int b);

        int minus(int a, int b);

        abstract static class Stub extends Binder implements ICalculator {//看3.2分析

            private static final int CODE_ADD = 1;

            private static final int CODE_MINUS = 2;

            @Override
            protected boolean onTransact(int code, @NonNull Parcel data, @Nullable Parcel reply, int flags) throws RemoteException {

                if (code == CODE_ADD) {
                    int a = data.readInt();
                    int b = data.readInt();
                    reply.writeInt(add(a, b));
                    return true;
                } else if (code == CODE_MINUS) {
                    int a = data.readInt();
                    int b = data.readInt();
                    reply.writeInt(minus(a, b));
                    return true;
                }

                return super.onTransact(code, data, reply, flags);
            }

            abstract public int add(int a, int b);

            abstract public int minus(int a, int b);

            public static class Proxy implements ICalculator {//看3.3分析

                private IBinder remote;

                public Proxy(IBinder remote) {
                    this.remote = remote;
                }

                @Override
                public int add(int a, int b) {
                    Parcel data = Parcel.obtain();
                    Parcel reply = Parcel.obtain();
                    data.writeInt(a);
                    data.writeInt(b);
                    remote.transact(CODE_ADD, data, reply, 0);
                    int result = reply.readInt();
                    return result;
                }

                @Override
                public int minus(int a, int b) {
                    Parcel data = Parcel.obtain();
                    Parcel reply = Parcel.obtain();
                    data.writeInt(a);
                    data.writeInt(b);
                    remote.transact(CODE_MINUS, data, reply, 0);
                    int result = reply.readInt();
                    return result;
                }
            }
        }

    }

3.1 ICalculator接口

看起来和aidl文件差不多

    interface ICalculator {
        int add(int a, int b);
        int minus(int a, int b);
    }

3.2 ICalculator.Stub类

继承于Binder，实现ICalculator接口，但是是空实现，然后在onTransact方法中调用空实现的add和minus接口

  abstract static class Stub extends Binder implements ICalculator {

            private static final int CODE_ADD = 1;

            private static final int CODE_MINUS = 2;

            @Override
            protected boolean onTransact(int code, @NonNull Parcel data, @Nullable Parcel reply, int flags) throws RemoteException {

                if (code == CODE_ADD) {
                    int a = data.readInt();
                    int b = data.readInt();
                    reply.writeInt(add(a, b));
                    return true;
                } else if (code == CODE_MINUS) {
                    int a = data.readInt();
                    int b = data.readInt();
                    reply.writeInt(minus(a, b));
                    return true;
                }

                return super.onTransact(code, data, reply, flags);
            }

            abstract public int add(int a, int b);

            abstract public int minus(int a, int b);

        }

3.3 ICalculator.Stub.Proxy类

用BinderProxy构造ICalculator.Stub.Proxy，把调用ICalculator.Stub.Proxy的add和minus接口转化成Parcel.write,read和BinderProxy.transact代码。

          public static class Proxy implements ICalculator {

                private IBinder remote;

                public Proxy(IBinder remote) {
                    this.remote = remote;
                }

                @Override
                public int add(int a, int b) {
                    Parcel data = Parcel.obtain();
                    data.writeInt(a);
                    data.writeInt(b);
                    Parcel reply = Parcel.obtain();
                    remote.transact(CODE_ADD, data, reply, 0);
                    int result = reply.readInt();
                    return result;
                }

                @Override
                public int minus(int a, int b) {
                    Parcel data = Parcel.obtain();
                    data.writeInt(a);
                    data.writeInt(b);
                    Parcel reply = Parcel.obtain();
                    remote.transact(CODE_MINUS, data, reply, 0);
                    int result = reply.readInt();
                    return result;
                }
            }

总结

看完应该明白了AIDL作用主要有
a.偷懒，少很多代码，尤其是你要写上百个方法的时候，AIDL就派上大用场了。
b.让服务端更专注接口的实现，而减少犯错误的可能性
c.规范client端和server端的接口定义，有助于代码的迭代

当然这是我自己写的伪代码，AIDL真正生成的代码更加复杂一点，但这些伪代码就是AIDL核心点。

思考

oneway的这个语法对AIDL生成的代码有什么影响，或者你们自己写一下oneway的方法AIDL生成的伪代码