问题描述:
最近在写一个通信相关的项目,服务器端和客户端通过socket进行通信。本来想利用read的阻塞特性,服务器端和客户端按照一定的流程进行文件读写。结果发现客户端或者服务器read方法一直都返回乱码。而且读到的一端可能是客户端,可能是服务器端,固定的读到前面有四个字节的乱码,后续读到的字节码都是正常的。
原因分析:
开始以为是流没有正常关闭。修改了代码确保正确关闭之后,发现即使重新启动服务器和客户端,还是会固定读到四个字节乱码。后面查资料分析才找出真正的原因:由于我实现的socket通信既有字符串通信,又有对象通信。所以我在传递字符串的时候,使用的是socket.getOutputStream得到的流。而在进行对象传输的时候,我在前面的输出流外面包裹了一层ObjectOutputStream。因为我是在一开始就对socket的输出流进行了包裹,而如果用ObjectOutputStream装饰输出流,默认的会自动在流前面带上四个字节的前缀。而因为开始我发消息只是发送字符串,所以我是直接使用socket的输出流。这就导致将前面的四个字节前缀发送出去,导致最终的乱码。具体参见下面相关代码:
1 /**
2 * Creates an ObjectOutputStream that writes to the specified OutputStream.
3 * This constructor writes the serialization stream header to the
4 * underlying stream; callers may wish to flush the stream immediately to
5 * ensure that constructors for receiving ObjectInputStreams will not block
6 * when reading the header.
7 *
8 * <p>If a security manager is installed, this constructor will check for
9 * the "enableSubclassImplementation" SerializablePermission when invoked
10 * directly or indirectly by the constructor of a subclass which overrides
11 * the ObjectOutputStream.putFields or ObjectOutputStream.writeUnshared
12 * methods.
13 *
14 * @param out output stream to write to
15 * @throws IOException if an I/O error occurs while writing stream header
16 * @throws SecurityException if untrusted subclass illegally overrides
17 * security-sensitive methods
18 * @throws NullPointerException if <code>out</code> is <code>null</code>
19 * @since 1.4
20 * @see ObjectOutputStream#ObjectOutputStream()
21 * @see ObjectOutputStream#putFields()
22 * @see ObjectInputStream#ObjectInputStream(InputStream)
23 */
24 public ObjectOutputStream(OutputStream out) throws IOException {
25 verifySubclass();
26 bout = new BlockDataOutputStream(out);
27 handles = new HandleTable(10, (float) 3.00);
28 subs = new ReplaceTable(10, (float) 3.00);
29 enableOverride = false;
30 writeStreamHeader();
31 bout.setBlockDataMode(true);
32 if (extendedDebugInfo) {
33 debugInfoStack = new DebugTraceInfoStack();
34 } else {
35 debugInfoStack = null;
36 }
37 }
38
39
40
41 /**
42 * The writeStreamHeader method is provided so subclasses can append or
43 * prepend their own header to the stream. It writes the magic number and
44 * version to the stream.
45 *
46 * @throws IOException if I/O errors occur while writing to the underlying
47 * stream
48 */
49 protected void writeStreamHeader() throws IOException {
50 bout.writeShort(STREAM_MAGIC);
51 bout.writeShort(STREAM_VERSION);
52 }
解决办法:
既然直接用ObjectOutputStream将原来的socket的输出流进行包裹之后会出现固定四个字节的乱码,那么可以考虑用原来的socket输出流进行写数据的时候,接收方固定丢弃四个字节乱码。这样虽然可以实现,但是总感觉很别扭。最终我优化了相关的读写对象方法,只是用原来socket的输出流进行对象读写,具体代码实现如下:
1 public <T> void writeObj(T obj) throws ZSocketException {
2 if (obj == null) {
3 return;
4 }
5 try(ByteArrayOutputStream byteOut = new ByteArrayOutputStream();
6 ObjectOutputStream objOut = new ObjectOutputStream(byteOut)) { // 这个只是为了计算出对象大小而用的中介输出流
7 objOut.writeObject(obj);
8 byte[] ObjByte = byteOut.toByteArray();
9 Header header = new Header(StringMsgType.OBJECT, ObjByte.length);
10 HeaderAnalyser analyser = new HeaderAnalyser();
11
12 // 先写消息头,再写消息内容
13 output.write(analyser.formatHeader(header), 0, Constants.HEADER_LEN);
14 output.write(ObjByte, 0, ObjByte.length);
15 output.flush();
16
17 } catch (IOException e) {
18 throw new ZSocketException(e);
19 }
20 }
21
22 public <T> T readObj(long len, Class<T> clazz) throws ZSocketException {
23 if (len < 0 || clazz == null) {
24 throw new IllegalArgumentException("Negative read length or null object class!");
25 }
26
27 try(ByteArrayOutputStream out = new ByteArrayOutputStream(Constants.BUFF_SIZE)) {
28 writeData(input, out, len);
29 try (ByteArrayInputStream byteIn = new ByteArrayInputStream(out.toByteArray());
30 ObjectInputStream objIn = new ObjectInputStream(byteIn)) {
31 @SuppressWarnings("unchecked")
32 T result = (T) objIn.readObject();
33 return result;
34 }
35 } catch (Exception e) {
36 throw new ZSocketException(e);
37 }
38
39 }