javascript Node.js TCP 服务器传入缓冲区
声明：本页面是StackOverFlow热门问题的中英对照翻译，遵循CC BY-SA 4.0协议，如果您需要使用它，必须同样遵循CC BY-SA许可，注明原文地址和作者信息，同时你必须将它归于原作者（不是我）：StackOverFlow
原文地址: http://stackoverflow.com/questions/5955498/
Warning: these are provided under cc-by-sa 4.0 license. You are free to use/share it, But you must attribute it to the original authors (not me):
StackOverFlow
Node.js TCP server incoming buffer
提问by user747454
I have two node processes that speak to each other. I will call them [Node Server]and [Node Sender]. [Node Sender]continually processes information and writes a message over a TCP connection to [Node Server]. [Node Server]then writes back a status message.
我有两个相互通信的节点进程。我将它们称为[Node Server]和[Node Sender]。 [Node Sender]不断处理信息，并通过 TCP 连接将消息写入[Node Server]。 [节点服务器]然后写回状态消息。
Example of [Node Sender]:
[节点发送者]示例：
var message = "Test Message";
[Node Sender].connection.write(message);
Example of [Node Server]:
【节点服务器】示例：
[Node Server].socket.on("data", function(p_data) {
this.write("OK");
// Do some work with p_data
}
This works without issue, p_dataalways contains "Test Message"when sent at anything above 5 milliseconds. However, if I speed [Node Sender]to write every millisecond, p_dataoccasionally ends up with something like "Test MessageTest MessageTes".
这可以正常工作，p_data在以 5 毫秒以上的任何时间发送时始终包含“测试消息”。但是，如果我加速[Node Sender]每毫秒写入一次，p_data偶尔会以类似"Test MessageTest MessageTes" 的方式结束。
I understand that the buffer in [Node Sender]is probably filling faster than the write command is sending it. Is there a way to force a one-to-one ratio in sending messages while still remaining asynchronous?
我知道[Node Sender]中的缓冲区填充速度可能比写入命令发送它的速度快。有没有办法强制一对一的消息发送同时保持异步？
I can certainly just add a terminator to my message and just fill a buffer in [Node Server], but I wanted to make sure there wasn't something obvious I was missing.
我当然可以在我的消息中添加一个终止符并在[Node Server] 中填充一个缓冲区，但我想确保没有明显我遗漏的东西。
回答by Rob Raisch
No, you're not missing anything and yes, you do need to add some form of termination to your messages.
不，您没有遗漏任何东西，是的，您确实需要在消息中添加某种形式的终止。
There are two basic problems here:
这里有两个基本问题：
The TCP protocol is stream-oriented, not message-oriented; it has no intrinsic knowledge of what might constitute a "message".
The data event fired by the node.js net library indicates that somedata has arrived but without any idea of what a message might contain, it cannot indicate that it has received any specific data.
TCP 协议是面向流的，而不是面向消息的；它对可能构成"消息"的内容没有内在知识。
node.js 网络库触发的数据事件表明一些数据已经到达，但不知道消息可能包含什么，它不能表明它已收到任何特定数据。
So by sending your messages faster than Node can process them, the socket recv buffer fills with multiple "messages".
因此，通过以比 Node 处理它们更快的速度发送您的消息，套接字接收缓冲区会填充多个“消息”。
A typical solution to this problem is to add line-termination, as can be found in https://github.com/baudehlo/Haraka/blob/master/connection.jsat lines 32-34:
此问题的典型解决方案是添加行终止，如 https://github.com/baudehlo/Haraka/blob/master/connection.js第 32-34 行所示：
self.client.on('data', function (data) {
self.process_data(data);
});
and lines 110-134:
和第 110-134 行：
Connection.prototype.process_data = function (data) {
if (this.disconnected) {
logger.logwarn("data after disconnect from " + this.remote_ip);
return;
}
this.current_data += data;
this._process_data();
};
Connection.prototype._process_data = function() {
var results;
while (results = line_regexp.exec(this.current_data)) {
var this_line = results[1];
if (this.state === 'pause') {
this.early_talker = 1;
var self = this;
// If you talk early, we're going to give you a delay
setTimeout(function() { self._process_data() }, this.early_talker_delay);
break;
}
this.current_data = this.current_data.slice(this_line.length);
this.process_line(this_line);
}
};
回答by jeremyko
You need to accumulate incoming buffer data to get your complete message. please refer to below example. this server expects data with 4 byte header and message body. header is unsigned int which means total length of body and the body is string data with delimiter '|'. please note that it is possible that this 'header and message' is not being received at a time. so we need to accumulate incoming data until we get a full length of data. And it is also possible that multiple 'header and message' is being received at a time. The point is that we need to data accumulation.
您需要累积传入的缓冲区数据才能获得完整的消息。请参考下面的例子。该服务器需要具有 4 字节标头和消息正文的数据。标头是 unsigned int，这意味着主体的总长度，主体是带有分隔符“|”的字符串数据。请注意，有可能一次未收到此“标题和消息”。所以我们需要积累传入的数据，直到我们得到完整长度的数据。并且一次接收多个“标题和消息”也是可能的。关键是我们需要数据积累。
var SERVER_PORT = 8124;
var TCP_DELIMITER = '|';
var packetHeaderLen = 4; // 32 bit integer --> 4
var server = net.createServer( function(c) {
var accumulatingBuffer = new Buffer(0);
var totalPacketLen = -1;
var accumulatingLen = 0;
var recvedThisTimeLen= 0;
var remoteAddress = c.remoteAddress;
var address= c.address();
var remotePort= c.remotePort;
var remoteIpPort = remoteAddress +":"+ remotePort;
console.log('-------------------------------'+remoteAddress);
console.log('remoteIpPort='+ remoteIpPort);
c.on('data', function(data) {
console.log('received data length :' + data.length );
console.log('data='+ data);
recvedThisTimeLen = data.length;
console.log('recvedThisTimeLen='+ recvedThisTimeLen);
//accumulate incoming data
var tmpBuffer = new Buffer( accumulatingLen + recvedThisTimeLen );
accumulatingBuffer.copy(tmpBuffer);
data.copy ( tmpBuffer, accumulatingLen ); // offset for accumulating
accumulatingBuffer = tmpBuffer;
tmpBuffer = null;
accumulatingLen += recvedThisTimeLen ;
console.log('accumulatingBuffer = ' + accumulatingBuffer );
console.log('accumulatingLen =' + accumulatingLen );
if( recvedThisTimeLen < packetHeaderLen ) {
console.log('need to get more data(less than header-length received) -> wait..');
return;
} else if( recvedThisTimeLen == packetHeaderLen ) {
console.log('need to get more data(only header-info is available) -> wait..');
return;
} else {
console.log('before-totalPacketLen=' + totalPacketLen );
//a packet info is available..
if( totalPacketLen < 0 ) {
totalPacketLen = accumulatingBuffer.readUInt32BE(0) ;
console.log('totalPacketLen=' + totalPacketLen );
}
}
//while=>
//in case of the accumulatingBuffer has multiple 'header and message'.
while( accumulatingLen >= totalPacketLen + packetHeaderLen ) {
console.log( 'accumulatingBuffer= ' + accumulatingBuffer );
var aPacketBufExceptHeader = new Buffer( totalPacketLen ); // a whole packet is available...
console.log( 'aPacketBufExceptHeader len= ' + aPacketBufExceptHeader.length );
accumulatingBuffer.copy( aPacketBufExceptHeader, 0, packetHeaderLen, accumulatingBuffer.length); //
////////////////////////////////////////////////////////////////////
//process one packet data
var stringData = aPacketBufExceptHeader.toString();
var usage = stringData.substring(0,stringData.indexOf(TCP_DELIMITER));
console.log("usage: " + usage);
//call handler
(serverFunctions [usage])(c, remoteIpPort, stringData.substring(1+stringData.indexOf(TCP_DELIMITER)));
////////////////////////////////////////////////////////////////////
//rebuild buffer
var newBufRebuild = new Buffer( accumulatingBuffer.length );
newBufRebuild.fill();
accumulatingBuffer.copy( newBufRebuild, 0, totalPacketLen + packetHeaderLen, accumulatingBuffer.length );
//init
accumulatingLen -= (totalPacketLen +4) ;
accumulatingBuffer = newBufRebuild;
newBufRebuild = null;
totalPacketLen = -1;
console.log( 'Init: accumulatingBuffer= ' + accumulatingBuffer );
console.log( ' accumulatingLen = ' + accumulatingLen );
if( accumulatingLen <= packetHeaderLen ) {
return;
} else {
totalPacketLen = accumulatingBuffer.readUInt32BE(0) ;
console.log('totalPacketLen=' + totalPacketLen );
}
}
});
...
});
please refer to below for whole example.
请参阅下面的完整示例。
https://github.com/jeremyko/nodeChatServer
https://github.com/jeremyko/nodeChatServer
Hope this help.
希望这有帮助。
