我有一个CompositeByteBuf，其中包含一些构成HTTP请求的缓冲区，我想在HTTP请求行之后插入一个额外的HTTP标头字段。 （我不想使用整个HTTP编码器/解码器，因为我只是代理数据，不需要将所有数据解析为HTTP）。
我该如何使用派生的缓冲区执行此操作，因此避免复制CompositeByteBuf的内容。我使用slice和readSlice进行的每一次尝试都产生indexoutofbounds错误或Stack Overflow。任何人都可以提出以下替代方案，而不需要复制整个CompositeBytebuf吗？
/**
* Injects an XFF header into pendingBuf
*/
private void addXForwardedForHeaderToPendingBuf(
int pLFpos,
String pRemoteIPaddr)
{
//create a new buffer
ByteBuf newBuf = inboundChannel.alloc().directBuffer();
//add the HTTP request line to it
ByteBufUtil.writeUtf8(newBuf,
pendingBuf.readCharSequence(pLFpos + 1,
CharsetUtil.UTF_8));
//add the XFF header
ByteBufUtil.writeUtf8(newBuf, "X-Forwarded-For: ");
ByteBufUtil.writeUtf8(newBuf, pRemoteIPaddr);
ByteBufUtil.writeUtf8(newBuf, "\r\n");
//add anything from the original buffer that came after the request line
int bytesRemaining = pendingBuf.readableBytes();
if (bytesRemaining > 0)
{
newBuf.writeBytes(pendingBuf);
}
//clear pendingBuf
pendingBuf.removeComponents(0, pendingBuf.numComponents());
pendingBuf.setIndex(0, 0);
//add newBuf into pendingBuf
pendingBuf.addComponent(newBuf);
pendingBuf.writerIndex(pendingBuf.writerIndex() + newBuf.writerIndex());
}
最佳答案
虽然编辑当前的字节缓冲区有一个缺点，那就是在最坏的情况下所有字节都需要移动，我们可以利用CompositeByteBuf具有可以编辑并根据需要移动的组件这一事实。
我们基本上想实现以下步骤：
由于Bytebuf中可能有多个CompositeByteBuf，因此我们要搜索要修改的buf的索引。ByteBuf为我们提供了以下方法：
toComponentIndex
toByteIndex
遗憾的是，这些方法在字符串末尾插入的情况下将无法正常工作，因为从技术上讲，这对于原始缓冲区是超出范围的，因此我们需要为此添加一个特殊情况。
我们想要实现一种特殊情况，即我们想精确地插入多个缓冲区之间的边界，因为在这些情况下，我们实际上可以使用零拷贝。
如果拆分索引恰好位于一个字节缓冲区的中间，则需要对其进行拆分，并将其自身添加为2个单独的缓冲区。
我们需要更新组合for some reason this doesn't happen by default.上的writer索引
使用以上流程，我们可以创建以下代码：
public static void insertString(CompositeByteBuf buffer, int index, ByteBuf insertion) {
try {
if (buffer == null) {
throw new NullPointerException("buffer");
}
if (insertion == null) {
throw new NullPointerException("insertion");
}
if (buffer.readableBytes() < index) {
throw new IllegalArgumentException("buffer.readableBytes() < index: "
+ buffer.readableBytes() + " < " + index);
}
// Start by checking the offset where we need to inject the insertion
int injectionBufOffset;
int injectionByteOffset;
if (index == buffer.readableBytes()) {
injectionBufOffset = buffer.numComponents();
injectionByteOffset = 0;
} else {
injectionBufOffset = buffer.toComponentIndex(index);
injectionByteOffset = index - buffer.toByteIndex(injectionBufOffset);
}
// Optimalize in the case of offset 0
if (injectionByteOffset == 0) {
buffer.addComponent(injectionBufOffset, insertion.retain());
buffer.writerIndex(buffer.writerIndex() + insertion.readableBytes());
return;
}
// Do the split technique
ByteBuf toSplit = buffer.internalComponent(injectionBufOffset).retain();
try {
buffer.removeComponent(injectionBufOffset);
buffer.addComponent(injectionBufOffset + 0,
toSplit.readSlice(injectionByteOffset).retain());
buffer.addComponent(injectionBufOffset + 1, insertion.retain());
buffer.addComponent(injectionBufOffset + 2,
toSplit.retain());
buffer.writerIndex(buffer.writerIndex() + insertion.readableBytes());
} finally {
ReferenceCountUtil.release(toSplit);
}
} finally {
if (insertion != null) {
ReferenceCountUtil.release(insertion);
}
}
}
由于此代码非常复杂，因此我们还想确保对其进行了正确的测试，因此，我们需要一些单元测试（JUnit）：
import static test.NettySplit.insertString;
public class NettySplitTest {
CompositeByteBuf buffer;
ByteBuf test;
private void addByteBuf(CompositeByteBuf target, ByteBuf source) {
target.addComponent(source);
target.writerIndex(target.writerIndex() + source.readableBytes());
}
@Before
public void before() {
buffer = ByteBufAllocator.DEFAULT.compositeBuffer();
}
@After
public void after() {
ReferenceCountUtil.release(buffer);
buffer = null;
ReferenceCountUtil.release(test);
test = null;
}
@Test
public void testSplitting() {
addByteBuf(buffer, Unpooled.wrappedBuffer(new byte[]{0, 1, 2, 3}));
insertString(buffer, 2, Unpooled.wrappedBuffer(new byte[]{5}));
test = Unpooled.wrappedBuffer(new byte[]{0, 1, 5, 2, 3});
assertEquals(test, buffer);
}
@Test
public void testInsertionStart() {
addByteBuf(buffer, Unpooled.wrappedBuffer(new byte[]{0, 1, 2, 3}));
insertString(buffer, 0, Unpooled.wrappedBuffer(new byte[]{5}));
test = Unpooled.wrappedBuffer(new byte[]{5, 0, 1, 2, 3});
assertEquals(test, buffer);
}
@Test
public void testInsertionEnd() {
addByteBuf(buffer, Unpooled.wrappedBuffer(new byte[]{0, 1, 2, 3}));
insertString(buffer, 4, Unpooled.wrappedBuffer(new byte[]{5}));
test = Unpooled.wrappedBuffer(new byte[]{0, 1, 2, 3, 5});
assertEquals(test, buffer);
}
@Test
public void testInsertionSplitEnd() {
addByteBuf(buffer, Unpooled.wrappedBuffer(new byte[]{0, 1, 2, 3}));
addByteBuf(buffer, Unpooled.wrappedBuffer(new byte[]{0, 1, 2, 3}));
insertString(buffer, 6, Unpooled.wrappedBuffer(new byte[]{5}));
test = Unpooled.wrappedBuffer(new byte[]{0, 1, 2, 3, 0, 1, 5, 2, 3});
assertEquals(test, buffer);
}
}
关于java - 如何在Netty CompositeByteBuf的中间插入字符串，我们在Stack Overflow上找到一个类似的问题：https://stackoverflow.com/questions/44678122/