摘要：apache thrift作为著名的跨语言，跨平台的RPC框架已经得到了大量应用，比如Hadoop, Cassandra等。与早期的CORBA, DCOM, 以致传统的WebService如基于XML-RPC的 SOAP协议和基于http的restfull ws相比较，其强大的功能和性能以及开发效率都让人侧目。本文目的在于在windows上使用thrift编译一个RPC的java客户端和服务器来体验其使用。
关键字: RPC , thrift
1.下载和安装thrift
到 http://apache.fayea.com/thrift/0.9.2/thrift-0.9.2.exe 下载，可直接将下载目录加到Windows环境变量，以便可以随时敲thrift-0.9.2命令。
2.编译tutorial.thrift文件为java 代码：
到 https://git1-us-west.apache.org/repos/asf?p=thrift.git;a=tree;f=tutorial;hb=HEAD 下载tutorial.thrift和shared.thrift 2个文件 , 先执行 thrift-0.9.2 -gen java tutorial\shared.thrift 然后执行 thrift-0.9.2 -gen java tutorial\tutorial.thrift , 可以发现生成了gen-java目录，
将生成的java 类 进行编译会报错，放于eclipse工程中会发现很多包缺失，比如org.apache.thrift.* ;
org.apache.thrift报缺失是因为thrift的java类库需要构建出来。如果我们只下载了thrift-0.9.2.exe 那么是不带thrift-0.9.2.jar这个类库的。
可以下载thrift的源码 （http://mirrors.hust.edu.cn/apache/thrift/0.9.2/thrift-0.9.2.tar.gz）将其解压，到目录lib\java下执行ant (需要之前安装java和ant,并将ant folder\bin加到环境变量).会生成lib\java\build,其下有thrift-0.9.2.jar
添加CalculatorHandler.java :
import java.util.HashMap;
import tutorial.Calculator;
import tutorial.InvalidOperation;
import tutorial.SharedStruct;
import tutorial.Work;
public class CalculatorHandler implements Calculator.Iface {
private HashMap<Integer, SharedStruct> log;
public CalculatorHandler() {
log = new HashMap<Integer, SharedStruct>();
}
public void ping() {
System.out.println("ping()");
}
public int add(int n1, int n2) {
System.out.println("add(" + n1 + "," + n2 + ")");
return n1 + n2;
}
public int calculate(int logid, Work work) throws InvalidOperation {
System.out.println("calculate(" + logid + ", {" + work.op + ","
+ work.num1 + "," + work.num2 + "})");
int val = 0;
switch (work.op) {
case ADD:
val = work.num1 + work.num2;
break;
case SUBTRACT:
val = work.num1 - work.num2;
break;
case MULTIPLY:
val = work.num1 * work.num2;
break;
case DIVIDE:
if (work.num2 == 0) {
InvalidOperation io = new InvalidOperation();
io.what = work.op.getValue();
io.why = "Cannot divide by 0";
throw io;
}
val = work.num1 / work.num2;
break;
default:
InvalidOperation io = new InvalidOperation();
io.what = work.op.getValue();
io.why = "Unknown operation";
throw io;
}
SharedStruct entry = new SharedStruct();
entry.key = logid;
entry.value = Integer.toString(val);
log.put(logid, entry);
return val;
}
public SharedStruct getStruct(int key) {
System.out.println("getStruct(" + key + ")");
return log.get(key);
}
public void zip() {
System.out.println("zip()");
}
}
<strong>
</strong>import org.apache.thrift.TException;
import org.apache.thrift.protocol.TBinaryProtocol;
import org.apache.thrift.protocol.TProtocol;
import org.apache.thrift.transport.TSocket;
import org.apache.thrift.transport.TTransport;
import tutorial.Calculator;
import tutorial.InvalidOperation;
import tutorial.Operation;
import tutorial.SharedStruct;
import tutorial.Work;
public class Client {
static void simple() throws Exception {
TTransport transport = new TSocket("localhost", 9090);
transport.open();
TProtocol protocol = new TBinaryProtocol(transport);
Calculator.Client client = new Calculator.Client(protocol);
perform(client);
transport.close();
}
private static void perform(Calculator.Client client) throws TException
{
client.ping();
System.out.println("ping()");
int sum = client.add(1,1);
System.out.println("1+1=" + sum);
Work work = new Work();
work.op = Operation.DIVIDE;
work.num1 = 1;
work.num2 = 0;
try {
int quotient = client.calculate(1, work);
System.out.println("Whoa we can divide by 0");
} catch (InvalidOperation io) {
System.out.println("Invalid operation: " + io.why);
}
work.op = Operation.SUBTRACT;
work.num1 = 15;
work.num2 = 10;
try {
int diff = client.calculate(1, work);
System.out.println("15-10=" + diff);
} catch (InvalidOperation io) {
System.out.println("Invalid operation: " + io.why);
}
SharedStruct log = client.getStruct(1);
System.out.println("Check log: " + log.value);
}
/**
* @param args
* @throws Exception
*/
public static void main(String[] args) throws Exception {
simple();
}
}
import org.apache.thrift.server.TServer;
import org.apache.thrift.server.TServer.Args;
import org.apache.thrift.transport.TServerSocket;
import org.apache.thrift.transport.TServerTransport;
import org.apache.thrift.server.TSimpleServer;
import tutorial.Calculator;
public class Server {
static final int port_Simple = 9090;
public static void simple(Calculator.Processor processor) {
try {
// Args arg = null;
TServerTransport serverTransport = new TServerSocket(port_Simple);
TServer server = new TSimpleServer( new Args(serverTransport).processor(processor));
// Use this for a multithreaded server
// TServer server = new TThreadPoolServer(new TThreadPoolServer.Args(serverTransport).processor(processor));
System.out.println("Starting the simple server...");
server.serve();
} catch (Exception e) {
e.printStackTrace();
}
}
/**
* @param args
*/
public static void main(String[] args) {
CalculatorHandler handler = new CalculatorHandler();
Calculator.Processor processor = new Calculator.Processor(handler);
simple(processor);
}
}---引用jar: libthrift-0.9.2.jar, slf4j-*.jar
4.启动 server，再运行client可以看到客户端输出：
ping()
1+1=2
Invalid operation: Cannot divide by 0
15-10=5
Check log: 5
总结：
1. thrift缺点：
1)文档少，初学者要花较长时间来琢磨如何写出第一个基于thrift RPC的C/S程序.
2)文档bug多，文档描述与实际运行的行为不符。比如根据文档的tutorial/tutorial.thrift来生成Java代码时 并没有生成shared.SharedService类。
3)windows上的thrift编译器(比如thrift-0.9.2.exe)有bug,对于thrift文件中的include指示符支持不好。
2. thrift优点：
1)跨平台，跨/多语言的支持非常好。比如一个运行在Windows上的c#客户端完全可以访问一个运行在Linux机器上的用Java实现的服务器端。
thrift编译器可以根据thrift文件编译出针对当下几乎所有流行语言的代码()。
2)性能好。 如果服务端代码用c/c++则性能更是杠杠的。
3)开发效率高，手工代码量小。 thrift编译器所生成的代码基本上拿来即可使用，（需要增加服务端代码来实现自己的业务逻辑，但是协议实现，数据类型实现都已经自动生成）。
4)兼容性好。如果thrift文件更改，那么之前基于老thrift文件的程序仍能良好运行。
参考资料：
1. Apache thrift网站：http://thrift.apache.org/
2. RPC性能比较： http://www.useopen.net/blog/2015/rpc-performance.html
3. 董西城的博客： http://dongxicheng.org/search-engine/thrift-rpc/