前言
在上一篇随笔中,我们探讨了如何实现一套自定义通信协议,其中涉及到的粘包和拆包处理最初是完全自定义实现的,后来则改为了继承 ByteToMessageDecoder
来简化处理。
本篇将重点讨论这两种实现方式在缓存管理上的主要区别,并深入分析其中的不同之处以及值得借鉴的经验和技巧。
代码回顾
1)完全自定义实现
无缓存的情况
- 反复从ByteBuf中提取完整的消息
- 剩余的残缺消息写入缓存(会进行数据拷贝)
有缓存的情况
- 将新收到的数据接入缓存
- 反复从缓存中提取完整消息
- 释放缓存内读取过的数据(会进行数据移动,导致拷贝)
public class EchoServerHandler extends ChannelInboundHandlerAdapter {
private static final int HEADER_LENGTH = 4; //消息头部长度
private ByteBuf buffer = Unpooled.buffer(1024); //缓存残缺消息
@Override
public void channelRead(ChannelHandlerContext ctx, Object msg) throws Exception {
ByteBuf income = (ByteBuf) msg;
//上一次有缓存存在,则本数据包不是消息头开头,
if(buffer.readableBytes() > 0) {
//进行必要的扩容,下面的readBytes不会自动扩容
buffer.ensureWritable(income.readableBytes());
income.readBytes(buffer, income.readableBytes());
readMsgFromBuffer(buffer);
//剩下一点残缺消息
if(buffer.readableBytes() > 0) {
//保留剩下的数据,重置读索引为0
System.out.println("缓存剩余字节:"+buffer.readableBytes());
buffer.discardReadBytes();
} else { //刚刚好,则清空数据
buffer.clear();
}
} else {
readMsgFromBuffer(income);
//剩下的数据全部写入缓存
if (income.readableBytes() >0) {
System.out.println("剩余字节:"+income.readableBytes());
income.readBytes(buffer, income.readableBytes());
}
}
}
//从字节数组中读取完整的消息
private void readMsgFromBuffer(ByteBuf byteBuf) {
//剩余可读消息是否包含一个消息头
while(byteBuf.readableBytes() >= HEADER_LENGTH) {
byteBuf.markReaderIndex(); //由于可能读不到完整的消息,所以读之前先标记索引位置,方便重置
//读取消息头
byte[] headerBytes = new byte[4];
byteBuf.readBytes(headerBytes);
//获取类型
int type = headerBytes[0] & 0xFF;
//获取消息体长度
int bodyLength = ((headerBytes[1] & 0xFF) << 16) |
((headerBytes[2] & 0xFF) << 8) |
(headerBytes[3] & 0xFF);
//不包含请求体
if (byteBuf.readableBytes() < bodyLength) {
byteBuf.resetReaderIndex(); //重置读索引到当前消息头位置
break;
}
// 完整消息体已经接收,处理消息
byte[] body = new byte[bodyLength];
byteBuf.readBytes(body);
//System.out.println("type:"+type+"||length:"+bodyLength+"||body:"+new String(body, CharsetUtil.UTF_8));
if(type == 1) {
try {
HelloRequest request = HelloRequest.parseFrom(body);
System.out.println("收到消息:"+request.toString());
} catch (Exception e) {
System.out.println("解析失败:"+new String(body, CharsetUtil.UTF_8));
}
} else {
System.out.println("消息类型未知:"+type);
}
}
}
....
}
2)继承ByteToMessageDecoder的实现
使用ByteToMessageDecoder后,数据的解码变得更加简化。只需检查缓冲区是否有足够的数据来提取一个/多个完整的消息。
如果数据不足,解码过程就会结束,无需额外管理缓存。
public class MessageDecoder extends ByteToMessageDecoder {
private static final int HEADER_LENGTH = 4; //消息头部长度
@Override
protected void decode(ChannelHandlerContext ctx, ByteBuf in, List<Object> out) throws Exception {
// 检查是否足够的字节来读取一个消息头
while (in.readableBytes() >= HEADER_LENGTH) {
in.markReaderIndex(); // 标记当前读取位置,便于重置
// 读取消息头部
byte[] headerBytes = new byte[4];
in.readBytes(headerBytes);
// 获取类型
int type = headerBytes[0] & 0xFF;
// 获取消息体长度
int bodyLength = ((headerBytes[1] & 0xFF) << 16) |
((headerBytes[2] & 0xFF) << 8) |
(headerBytes[3] & 0xFF);
// 检查缓冲区中的数据是否足够读取整个消息体
if (in.readableBytes() < bodyLength) {
in.resetReaderIndex(); // 重置读指针,等待更多数据
break;
}
// 读取消息体
byte[] body = new byte[bodyLength];
in.readBytes(body);
// 处理消息
try {
Object msg = null;
if(type == 1) {
msg = HelloRequest.parseFrom(body);
} else if(type == 2) {
msg = HelloResponse.parseFrom(body);
} else {
System.out.println("未知消息:"+new String(body, CharsetUtil.UTF_8));
}
if(Objects.nonNull(msg)) {
out.add(msg);
}
} catch (Exception e) {
System.out.println("解析失败: " + new String(body, CharsetUtil.UTF_8));
}
}
}
}
ByteToMessageDecoder源码
核心属性
//缓存
private ByteBuf cumulation;
//累加器(用于拼接缓存和新到数据)
private Cumulator cumulator = MERGE_CUMULATOR;
//X次channelRead之后,释放已读数据
private int discardAfterReads = 16;
//累计channelRead次数(每次释放完会重置)
private int numReads;
处理流程
1.新到数据存放到缓冲区(使用累加器Cumulator进行数据合并)
2.循环调用子类的decode方法,读取消息存入List,直到数据不足
3.遍历List,依次传递给下一个处理器
累加器
提供2种累加器实现,MERGE_CUMULATOR和COMPOSITE_CUMULATOR
1)MERGE_CUMULATOR(默认实现)
缓存存在的时候,直接进行数据拷贝,与缓存数据进行整合。
下面的代码可以看到,如果缓冲区空间不够,则会进行扩容操作。
跟自定义实现中的"buffer.ensureWritable(income.readableBytes())"一致。
整体思路跟自定义实现差不多,不过它多考虑了两种情况
- 数据被共享:共享数据会被其他使用者影响,需排除影响
- 数据只读:只读空间无法被写入,而缓冲区是需要写入新数据的
public static final Cumulator MERGE_CUMULATOR = new Cumulator() {
//cumulation是上一次的缓存,in是新到的数据
@Override
public ByteBuf cumulate(ByteBufAllocator alloc, ByteBuf cumulation, ByteBuf in) {
try {
final ByteBuf buffer;
if (cumulation.writerIndex() > cumulation.maxCapacity() - in.readableBytes()
|| cumulation.refCnt() > 1 || cumulation.isReadOnly()) {
// Expand cumulation (by replace it) when either there is not more room in the buffer
// or if the refCnt is greater then 1 which may happen when the user use slice().retain() or
// duplicate().retain() or if its read-only.
//
// See:
// - https://github.com/netty/netty/issues/2327
// - https://github.com/netty/netty/issues/1764
buffer = expandCumulation(alloc, cumulation, in.readableBytes());
} else {
buffer = cumulation;
}
//新到数据写入缓存
buffer.writeBytes(in);
return buffer;
} finally {
// We must release in in all cases as otherwise it may produce a leak if writeBytes(...) throw
// for whatever release (for example because of OutOfMemoryError)
in.release();
}
}
};
2)COMPOSITE_CUMULATOR
上面的处理,新到数据与缓存的合并是通过数据拷贝。而下面这种方式,则是使用组合(数据没有移动,只是提供一个整合后的视图)
public static final Cumulator COMPOSITE_CUMULATOR = new Cumulator() {
@Override
public ByteBuf cumulate(ByteBufAllocator alloc, ByteBuf cumulation, ByteBuf in) {
ByteBuf buffer;
try {
if (cumulation.refCnt() > 1) {
// Expand cumulation (by replace it) when the refCnt is greater then 1 which may happen when the
// user use slice().retain() or duplicate().retain().
//
// See:
// - https://github.com/netty/netty/issues/2327
// - https://github.com/netty/netty/issues/1764
buffer = expandCumulation(alloc, cumulation, in.readableBytes());
buffer.writeBytes(in);
} else {
CompositeByteBuf composite;
if (cumulation instanceof CompositeByteBuf) {
//上一次缓存已经是组合对象
composite = (CompositeByteBuf) cumulation;
} else {
composite = alloc.compositeBuffer(Integer.MAX_VALUE);
//缓存加入组合
composite.addComponent(true, cumulation);
}
//新到数据加入组合
composite.addComponent(true, in);
in = null;
buffer = composite;
}
return buffer;
} finally {
//由于使用组合方式,数据还在原来的地方。不能直接释放
if (in != null) {
// We must release if the ownership was not transferred as otherwise it may produce a leak if
// writeBytes(...) throw for whatever release (for example because of OutOfMemoryError).
in.release();
}
}
}
};
主要方法------channelRead
在上述的自定义实现中,每次从缓冲区读取完数据,会释放掉已读数据,防止缓存数据无限增长。
buffer.discardReadBytes();
而这里做了优化,累积16次读取后,才会进行释放。(channelReadComplete的时候也会触发)
这样做的好处,就是可以减少数据拷贝的次数。(discard操作会把已读数据清空,重置读索引,然后把剩余数据往前挪)
@Override
public void channelRead(ChannelHandlerContext ctx, Object msg) throws Exception {
//仅处理ByteBuf,其他消息直接传给下一个Handler
if (msg instanceof ByteBuf) {
CodecOutputList out = CodecOutputList.newInstance();
try {
ByteBuf data = (ByteBuf) msg;
first = cumulation == null;
//缓冲区为空,直接赋值
if (first) {
cumulation = data;
} else {
//使用累加器进行数据合并
cumulation = cumulator.cumulate(ctx.alloc(), cumulation, data);
}
//调用子类实现,从缓冲区中解析消息
callDecode(ctx, cumulation, out);
} catch (DecoderException e) {
throw e;
} catch (Exception e) {
throw new DecoderException(e);
} finally {
if (cumulation != null && !cumulation.isReadable()) {
//缓冲区数据刚好读完,清空缓冲区,清空已读次数
numReads = 0;
cumulation.release();
cumulation = null;
} else if (++ numReads >= discardAfterReads) {
// We did enough reads already try to discard some bytes so we not risk to see a OOME.
// See https://github.com/netty/netty/issues/4275
//已读数达到限定次数(默认16),释放已读数据
numReads = 0;
discardSomeReadBytes();
}
int size = out.size();
//是不是没解析到消息
decodeWasNull = !out.insertSinceRecycled();
//将解析出来的消息逐个传个下一个Handler
fireChannelRead(ctx, out, size);
//清空List,下次再用
out.recycle();
}
} else {
//直接丢给下一个Handler
ctx.fireChannelRead(msg);
}
}
主要方法------callDecode
这里主要通过检查List结果集和数据读取情况,来判断要不要结束解码循环。
protected void callDecode(ChannelHandlerContext ctx, ByteBuf in, List<Object> out) {
try {
while (in.isReadable()) {
//先读取List大小
int outSize = out.size();
//有数据,则先传给下一个Handler
if (outSize > 0) {
fireChannelRead(ctx, out, outSize);
out.clear();
// Check if this handler was removed before continuing with decoding.
// If it was removed, it is not safe to continue to operate on the buffer.
//
// See:
// - https://github.com/netty/netty/issues/4635
if (ctx.isRemoved()) {
break;
}
outSize = 0;
}
//开始之前,先记录可读数据量
int oldInputLength = in.readableBytes();
//调用子类decode方法
decodeRemovalReentryProtection(ctx, in, out);
// Check if this handler was removed before continuing the loop.
// If it was removed, it is not safe to continue to operate on the buffer.
//
// See https://github.com/netty/netty/issues/1664
if (ctx.isRemoved()) {
break;
}
//查看子类是否解析出数据
if (outSize == out.size()) {
//数据没被动过,说明没有可解析的数据,直接break
if (oldInputLength == in.readableBytes()) {
break;
} else { //数据有被动过,但还没解析出数据,继续执行
continue;
}
}
//List内有新数据,但是数据没有被读过,说明子类实现有问题,报错
if (oldInputLength == in.readableBytes()) {
throw new DecoderException(
StringUtil.simpleClassName(getClass()) +
".decode() did not read anything but decoded a message.");
}
//如果只解析一次,则直接结束
if (isSingleDecode()) {
break;
}
}
} catch (DecoderException e) {
throw e;
} catch (Exception cause) {
throw new DecoderException(cause);
}
}
总结
核心内容并无太大差异,但 Netty 提供的抽象类在实现上考虑了更多细节,并经过社区的不断演进,功能变得更加稳定和完善。
因此,推荐继承 ByteToMessageDecoder
来实现解码。
其中,减少释放次数的设计思想值得学习。