我像这样创建一个新的双工流
const Duplex = require('stream').Duplex;
let myStream = new Duplex()
通过Websocket，我收到块/缓冲区，每次通过Websocket传入新块时，我都会像这样将它们添加到流中：
myStream.push(buffer)
然后我将流传输到另一个进程（在此示例中为ffmpeg）
myStream.pipe(process.stdout);
这会导致我理解的错误NodeError: The _read() method is not implemented，但我不明白为什么以及如何实现它。我还看到在Duplex类构造函数中，您可以传递read函数，但是为什么这是必需的呢？我只是想不断将大块推入流中，然后将其通过管道传送到另一个进程
nodejs Duplex流要求实现者同时指定write和read方法：
import stream from 'stream';
const duplex = new stream.Duplex({
write: (chunk,encoding,next) {
// Do something with the chunk and then call next() to indicate
// that the chunk has been processed. The write() fn will handle
// data piped into this duplex stream. After the write() has
// finished,the data will be processed by the read() below.
next();
},read: ( size ) {
// Add new data to be read by streams piped from this duplex
this.push( "some data" )
}
})
有关流的官方nodejs文档可在这里找到：API for Stream Implementers
网络套接字方案
上述的websocket示例可能应该使用Readable而不是双工流。双工流在存储转发或处理转发方案中很有用。但是，听起来好像websocket示例中的流仅用于将数据从websocket移动到流接口。这可以使用Readable来实现：
import stream from 'stream';
const onSocketConnection = ( socket ) => {
const readable = new stream.Readable({
// The read logic is omitted since the data is pushed to the socket
// outside of the script's control. However,the read() function
// must be defined.
read(){}
});
socket.on('message',( data ) => {
// Push the data on the readable queue
readable.push( data );
});
readable.pipe( ffmpeg );
}