Thrift源码分析（二）-- 协议和编解码

Thrift源码分析（二）-- 协议和编解码
协议和编解码是一个网络应用程序的核心问题之一，客户端和服务器通过约定的协议来传输消息(数据)，通过特定的格式来编解码字节流，并转化成业务消息，提供给上层框架调用。

Thrift的协议比较简单，它把协议和编解码整合在了一起。抽象类TProtocol定义了协议和编解码的顶层接口。个人感觉采用抽象类而不是接口的方式来定义顶层接口并不好，TProtocol关联了一个TTransport传输对象，而不是提供一个类似getTransport()的接口，导致抽象类的扩展性比接口差。

TProtocol主要做了两个事情:

1. 关联TTransport对象

2.定义一系列读写消息的编解码接口，包括两类，一类是复杂数据结构比如readMessageBegin, readMessageEnd, writeMessageBegin, writMessageEnd.还有一类是基本数据结构，比如readI32, writeI32, readString, writeString
[java] view plain copy
1. public abstract class TProtocol {
3. /**
4. * Transport
5. */
6. protected TTransport trans_;
8. 　public abstract void writeMessageBegin(TMessage message) throws TException;
10. public abstract void writeMessageEnd() throws TException;
12. public abstract void writeStructBegin(TStruct struct) throws TException;
14. public abstract void writeStructEnd() throws TException;
16. public abstract void writeFieldBegin(TField field) throws TException;
18. public abstract void writeFieldEnd() throws TException;
20. public abstract void writeFieldStop() throws TException;
22. public abstract void writeMapBegin(TMap map) throws TException;
24. public abstract void writeMapEnd() throws TException;
26. public abstract void writeListBegin(TList list) throws TException;
28. public abstract void writeListEnd() throws TException;
30. public abstract void writeSetBegin(TSet set) throws TException;
32. public abstract void writeSetEnd() throws TException;
34. public abstract void writeBool(boolean b) throws TException;
36. public abstract void writeByte(byte b) throws TException;
38. public abstract void writeI16(short i16) throws TException;
40. public abstract void writeI32(int i32) throws TException;
42. public abstract void writeI64(long i64) throws TException;
44. public abstract void writeDouble(double dub) throws TException;
46. public abstract void writeString(String str) throws TException;
48. public abstract void writeBinary(ByteBuffer buf) throws TException;
50. /**
51. * Reading methods.
52. */
54. public abstract TMessage readMessageBegin() throws TException;
56. public abstract void readMessageEnd() throws TException;
58. public abstract TStruct readStructBegin() throws TException;
60. public abstract void readStructEnd() throws TException;
62. public abstract TField readFieldBegin() throws TException;
64. public abstract void readFieldEnd() throws TException;
66. public abstract TMap readMapBegin() throws TException;
68. public abstract void readMapEnd() throws TException;
70. public abstract TList readListBegin() throws TException;
72. public abstract void readListEnd() throws TException;
74. public abstract TSet readSetBegin() throws TException;
76. public abstract void readSetEnd() throws TException;
78. public abstract boolean readBool() throws TException;
80. public abstract byte readByte() throws TException;
82. public abstract short readI16() throws TException;
84. public abstract int readI32() throws TException;
86. public abstract long readI64() throws TException;
88. public abstract double readDouble() throws TException;
90. public abstract String readString() throws TException;
92. public abstract ByteBuffer readBinary() throws TException;
94. /**
95. * Reset any internal state back to a blank slate. This method only needs to
96. * be implemented for stateful protocols.
97. */
98. public void reset() {}
100. /**
101. * Scheme accessor
102. */
103. public Class<? extends IScheme> getScheme() {
104. return StandardScheme.class;
105. }
107. }
所谓协议就是客户端和服务器端约定传输什么数据，如何解析传输的数据。对于一个RPC调用的协议来说，要传输的数据主要有:

调用方

1. 方法的名称，包括类的名称和方法的名称

2. 方法的参数，包括类型和参数值

3.一些附加的数据，比如附件，超时事件，自定义的控制信息等等

返回方

1. 调用的返回码

2. 返回值

3.异常信息

从TProtocol的定义我们可以看出Thrift的协议约定如下事情:

1. 先writeMessageBegin表示开始传输消息了，写消息头。Message里面定义了方法名，调用的类型，版本号，消息seqId

2. 接下来是写方法的参数，实际就是写消息体。如果参数是一个类，就writeStructBegin

3. 接下来写字段，writeFieldBegin, 这个方法会写接下来的字段的数据类型和顺序号。这个顺序号是Thrfit对要传输的字段的一个编码，从１开始

4. 如果是一个集合就writeListBegin/writeMapBegin，如果是一个基本数据类型，比如int, 就直接writeI32

5. 每个复杂数据类型写完都调用writeXXXEnd，直到writeMessageEnd结束

6. 读消息时根据数据类型读取相应的长度

每个writeXXX都是采用消息头+消息体的方式。我们来看TBinaryProtocol的实现。

1. writeMessgeBegin方法写了消息头，包括4字节的版本号和类型信息，字符串类型的方法名，４字节的序列号seqId

2. writeFieldBegin，写了１个字节的字段数据类型，和2个字节字段的顺序号

3. writeI32，写了４个字节的字节数组

4. writeString,先写４字节消息头表示字符串长度，再写字符串字节

5. writeBinary,先写４字节消息头表示字节数组长度，再写字节数组内容

6.readMessageBegin时，先读４字节版本和类型信息，再读字符串，再读４字节序列号

7.readFieldBegin，先读1个字节的字段数据类型，再读2个字节的字段顺序号

8. readString时，先读４字节字符串长度，再读字符串内容。字符串统一采用UTF-8编码
[java] view plain copy
1. public void writeMessageBegin(TMessage message) throws TException {
2. if (strictWrite_) {
3. int version = VERSION_1 | message.type;
4. writeI32(version);
5. writeString(message.name);
6. writeI32(message.seqid);
7. } else {
8. writeString(message.name);
9. writeByte(message.type);
10. writeI32(message.seqid);
11. }
12. }
14. public void writeFieldBegin(TField field) throws TException {
15. writeByte(field.type);
16. writeI16(field.id);
17. }
19. private byte[] i32out = new byte[4];
20. public void writeI32(int i32) throws TException {
21. i32out[0] = (byte)(0xff & (i32 >> 24));
22. i32out[1] = (byte)(0xff & (i32 >> 16));
23. i32out[2] = (byte)(0xff & (i32 >> 8));
24. i32out[3] = (byte)(0xff & (i32));
25. trans_.write(i32out, 0, 4);
26. }
28. public void writeString(String str) throws TException {
29. try {
30. byte[] dat = str.getBytes("UTF-8");
31. writeI32(dat.length);
32. trans_.write(dat, 0, dat.length);
33. } catch (UnsupportedEncodingException uex) {
34. throw new TException("JVM DOES NOT SUPPORT UTF-8");
35. }
36. }
38. public void writeBinary(ByteBuffer bin) throws TException {
39. int length = bin.limit() - bin.position();
40. writeI32(length);
41. trans_.write(bin.array(), bin.position() + bin.arrayOffset(), length);
42. }
44. public TMessage readMessageBegin() throws TException {
45. int size = readI32();
46. if (size < 0) {
47. int version = size & VERSION_MASK;
48. if (version != VERSION_1) {
49. throw new TProtocolException(TProtocolException.BAD_VERSION, "Bad version in readMessageBegin");
50. }
51. return new TMessage(readString(), (byte)(size & 0x000000ff), readI32());
52. } else {
53. if (strictRead_) {
54. throw new TProtocolException(TProtocolException.BAD_VERSION, "Missing version in readMessageBegin, old client?");
55. }
56. return new TMessage(readStringBody(size), readByte(), readI32());
57. }
58. }
60. public TField readFieldBegin() throws TException {
61. byte type = readByte();
62. short id = type == TType.STOP ? 0 : readI16();
63. return new TField("", type, id);
64. }
66. public String readString() throws TException {
67. int size = readI32();
69. if (trans_.getBytesRemainingInBuffer() >= size) {
70. try {
71. String s = new String(trans_.getBuffer(), trans_.getBufferPosition(), size, "UTF-8");
72. trans_.consumeBuffer(size);
73. return s;
74. } catch (UnsupportedEncodingException e) {
75. throw new TException("JVM DOES NOT SUPPORT UTF-8");
76. }
77. }
79. return readStringBody(size);
80. }
TProtocol定义了基本的协议信息，包括传输什么数据，如何解析传输的数据的基本方法。

还存在一个问题，就是服务器端如何知道客户端发送过来的数据是怎么组合的，比如第一个字段是字符串类型，第二个字段是int。这个信息是在IDL生成客户端时生成的代码时提供了。Thrift生成的客户端代码提供了读写参数的方法，这两个方式是一一对应的，包括字段的序号，类型等等。客户端使用写参数的方法，服务器端使用读参数的方法。

关于IDL生成的客户端代码会在后面的文章具体描述。下面简单看一下自动生成的代码

1. 方法的调用从writeMessageBegin开始，发送了消息头信息

2. 写方法的参数，也就是写消息体。方法参数由一个统一的接口TBase描述，提供了read和write的统一接口。自动生成的代码提供了read, write方法参数的具体实现

3. 写完结束
[java] view plain copy
1. public void write_args(org.apache.thrift.protocol.TProtocol prot) throws org.apache.thrift.TException {
2. prot.writeMessageBegin(new org.apache.thrift.protocol.TMessage("handle", org.apache.thrift.protocol.TMessageType.CALL, 0));
3. handle_args args = new handle_args();
4. args.setIdentity(identity);
5. args.setUid(uid);
6. args.setSid(sid);
7. args.setType(type);
8. args.setMessage(message);
9. args.setParams(params);
10. args.write(prot);
11. prot.writeMessageEnd();
12. }
14. public interface TBase<T extends TBase<?,?>, F extends TFieldIdEnum> extends Comparable<T>, Serializable {
16. public void read(TProtocol iprot) throws TException;
18. public void write(TProtocol oprot) throws TException;
19. ｝
21. public static class handle_args <strong>implements org.apache.thrift.TBase</strong><handle_args, handle_args._Fields>, java.io.Serializable, Cloneable {
22. private static final org.apache.thrift.protocol.TStruct STRUCT_DESC = new org.apache.thrift.protocol.TStruct("handle_args");
24. private static final org.apache.thrift.protocol.TField IDENTITY_FIELD_DESC = new org.apache.thrift.protocol.TField("identity", org.apache.thrift.protocol.TType.STRING, (short)1);
25. private static final org.apache.thrift.protocol.TField UID_FIELD_DESC = new org.apache.thrift.protocol.TField("uid", org.apache.thrift.protocol.TType.I64, (short)2);
26. private static final org.apache.thrift.protocol.TField SID_FIELD_DESC = new org.apache.thrift.protocol.TField("sid", org.apache.thrift.protocol.TType.STRING, (short)3);
27. private static final org.apache.thrift.protocol.TField TYPE_FIELD_DESC = new org.apache.thrift.protocol.TField("type", org.apache.thrift.protocol.TType.I32, (short)4);
28. private static final org.apache.thrift.protocol.TField MESSAGE_FIELD_DESC = new org.apache.thrift.protocol.TField("message", org.apache.thrift.protocol.TType.STRING, (short)5);
29. private static final org.apache.thrift.protocol.TField PARAMS_FIELD_DESC = new org.apache.thrift.protocol.TField("params", org.apache.thrift.protocol.TType.MAP, (short)6);
31. private static final Map<Class<? extends IScheme>, SchemeFactory> schemes = new HashMap<Class<? extends IScheme>, SchemeFactory>();
32. static {
33. schemes.put(StandardScheme.class, new handle_argsStandardSchemeFactory());
34. schemes.put(TupleScheme.class, new handle_argsTupleSchemeFactory());
35. }
37. public String identity; // required
38. public long uid; // required
39. public String sid; // required
40. public int type; // required
41. public String message; // required
42. public Map<String,String> params; // required
44. /** The set of fields this struct contains, along with convenience methods for finding and manipulating them. */
45. public enum _Fields implements org.apache.thrift.TFieldIdEnum {
46. IDENTITY((short)1, "identity"),
47. UID((short)2, "uid"),
48. SID((short)3, "sid"),
49. TYPE((short)4, "type"),
50. MESSAGE((short)5, "message"),
51. PARAMS((short)6, "params");
52. ｝
54. //　　自动生成的写方法参数的方法，按照字段顺序写，给客户端代码使用
55. public void write(org.apache.thrift.protocol.TProtocol oprot, handle_args struct) throws org.apache.thrift.TException {
56. struct.validate();
58. oprot.writeStructBegin(STRUCT_DESC);
59. if (struct.identity != null) {
60. oprot.writeFieldBegin(IDENTITY_FIELD_DESC);
61. oprot.writeString(struct.identity);
62. oprot.writeFieldEnd();
63. }
64. oprot.writeFieldBegin(UID_FIELD_DESC);
65. oprot.writeI64(struct.uid);
66. oprot.writeFieldEnd();
67. if (struct.sid != null) {
68. oprot.writeFieldBegin(SID_FIELD_DESC);
69. oprot.writeString(struct.sid);
70. oprot.writeFieldEnd();
71. }
72. oprot.writeFieldBegin(TYPE_FIELD_DESC);
73. oprot.writeI32(struct.type);
74. oprot.writeFieldEnd();
75. if (struct.message != null) {
76. oprot.writeFieldBegin(MESSAGE_FIELD_DESC);
77. oprot.writeString(struct.message);
78. oprot.writeFieldEnd();
79. }
80. ｝
82. <pre name="code" class="java">//　　自动生成的读方法参数的方法，按照字段顺序读，给服务器端代码使用
public void read(org.apache.thrift.protocol.TProtocol iprot, handle_args struct) throws org.apache.thrift.TException {
        org.apache.thrift.protocol.TField schemeField;
        iprot.readStructBegin();
        while (true)
        {
          schemeField = iprot.readFieldBegin();
          if (schemeField.type == org.apache.thrift.protocol.TType.STOP) {
            break;
          }
          switch (schemeField.id) {
            case 1: // IDENTITY
              if (schemeField.type == org.apache.thrift.protocol.TType.STRING) {
                struct.identity = iprot.readString();
                struct.setIdentityIsSet(true);
              } else {
                org.apache.thrift.protocol.TProtocolUtil.skip(iprot, schemeField.type);
              }
              break;
            case 2: // UID
              if (schemeField.type == org.apache.thrift.protocol.TType.I64) {
                struct.uid = iprot.readI64();
                struct.setUidIsSet(true);
              } else {
                org.apache.thrift.protocol.TProtocolUtil.skip(iprot, schemeField.type);
              }
              break;
            case 3: // SID
              if (schemeField.type == org.apache.thrift.protocol.TType.STRING) {
                struct.sid = iprot.readString();
                struct.setSidIsSet(true);
              } else {
                org.apache.thrift.protocol.TProtocolUtil.skip(iprot, schemeField.type);
              }
              break;
            case 4: // TYPE
              if (schemeField.type == org.apache.thrift.protocol.TType.I32) {
                struct.type = iprot.readI32();
                struct.setTypeIsSet(true);
              } else {
                org.apache.thrift.protocol.TProtocolUtil.skip(iprot, schemeField.type);
              }
              break;
｝
相关阅读:
django报错问题解决
 django2-项目实战
 django1-环境搭建与开始
 python学习-模块与包（九）
逆水行舟，不进则退
 待学习规划内容
 越来越强大法则
 windows下安装scoop
pytest+allure生成测试报告
 Angular vscode 调试
原文地址：https://www.cnblogs.com/heapStark/p/9218394.html