Table of Contents

• Stream
  • API for Stream Consumers
    • Class: stream.Duplex
    • Class: stream.Readable
      • Event: 'close'
      • Event: 'data'
      • Event: 'end'
      • Event: 'error'
      • Event: 'readable'
      • readable.isPaused()
      • readable.pause()
      • readable.pipe(destination[, options])
      • readable.read([size])
      • readable.resume()
      • readable.setEncoding(encoding)
      • readable.unpipe([destination])
      • readable.unshift(chunk)
      • readable.wrap(stream)
    • Class: stream.Transform
    • Class: stream.Writable
      • Event: 'close'
      • Event: 'drain'
      • Event: 'error'
      • Event: 'finish'
      • Event: 'pipe'
      • Event: 'unpipe'
      • writable.cork()
      • writable.end([chunk][, encoding][, callback])
      • writable.setDefaultEncoding(encoding)
      • writable.uncork()
      • writable.write(chunk[, encoding][, callback])
  • API for Stream Implementors
    • Class: stream.Duplex
      • new stream.Duplex(options)
    • Class: stream.PassThrough
    • Class: stream.Readable
      • new stream.Readable([options])
      • readable._read(size)
      • readable.push(chunk[, encoding])
      • Example: A Counting Stream
      • Example: SimpleProtocol v1 (Sub-optimal)
    • Class: stream.Transform
      • new stream.Transform([options])
      • Events: 'finish' and 'end'
      • transform._flush(callback)
      • transform._transform(chunk, encoding, callback)
      • Example: SimpleProtocol parser v2
    • Class: stream.Writable
      • new stream.Writable([options])
      • writable._write(chunk, encoding, callback)
      • writable._writev(chunks, callback)
  • Simplified Constructor API
    • Duplex
    • Readable
    • Transform
    • Writable
  • Streams: Under the Hood
    • Buffering
    • Compatibility with Older Node.js Versions
    • Object Mode
    • stream.read(0)
    • stream.push('')

Stream#

Stability: 2 - Stable

A stream is an abstract interface implemented by various objects in Node.js. For example a request to an HTTP server is a stream, as is process.stdout. Streams are readable, writable, or both. All streams are instances of EventEmitter.

You can load the Stream base classes by doing require('stream'). There are base classes provided for Readable streams, Writable streams, Duplex streams, and Transform streams.

This document is split up into 3 sections:

1. The first section explains the parts of the API that you need to be aware of to use streams in your programs.
2. The second section explains the parts of the API that you need to use if you implement your own custom streams yourself. The API is designed to make this easy for you to do.
3. The third section goes into more depth about how streams work, including some of the internal mechanisms and functions that you should probably not modify unless you definitely know what you are doing.

API for Stream Consumers#

Streams can be either Readable, Writable, or both (Duplex).

All streams are EventEmitters, but they also have other custom methods and properties depending on whether they are Readable, Writable, or Duplex.

If a stream is both Readable and Writable, then it implements all of the methods and events. So, a Duplex or Transform stream is fully described by this API, though their implementation may be somewhat different.

It is not necessary to implement Stream interfaces in order to consume streams in your programs. If you are implementing streaming interfaces in your own program, please also refer to API for Stream Implementors.

Almost all Node.js programs, no matter how simple, use Streams in some way. Here is an example of using Streams in an Node.js program:

const http = require('http');

var server = http.createServer( (req, res) => {
  // req is an http.IncomingMessage, which is a Readable Stream
  // res is an http.ServerResponse, which is a Writable Stream

  var body = '';
  // we want to get the data as utf8 strings
  // If you don't set an encoding, then you'll get Buffer objects
  req.setEncoding('utf8');

  // Readable streams emit 'data' events once a listener is added
  req.on('data', (chunk) => {
    body += chunk;
  });

  // the end event tells you that you have entire body
  req.on('end', () => {
    try {
      var data = JSON.parse(body);
    } catch (er) {
      // uh oh!  bad json!
      res.statusCode = 400;
      return res.end(`error: ${er.message}`);
    }

    // write back something interesting to the user:
    res.write(typeof data);
    res.end();
  });
});

server.listen(1337);

// $ curl localhost:1337 -d '{}'
// object
// $ curl localhost:1337 -d '"foo"'
// string
// $ curl localhost:1337 -d 'not json'
// error: Unexpected token o

Class: stream.Duplex#

Duplex streams are streams that implement both the Readable and Writable interfaces.

Examples of Duplex streams include:

• TCP sockets
• zlib streams
• crypto streams

Class: stream.Readable#

The Readable stream interface is the abstraction for a source of data that you are reading from. In other words, data comes out of a Readable stream.

A Readable stream will not start emitting data until you indicate that you are ready to receive it.

Readable streams have two "modes": a flowing mode and a paused mode. When in flowing mode, data is read from the underlying system and provided to your program as fast as possible. In paused mode, you must explicitly call stream.read() to get chunks of data out. Streams start out in paused mode.

Note: If no data event handlers are attached, and there are no stream.pipe() destinations, and the stream is switched into flowing mode, then data will be lost.

You can switch to flowing mode by doing any of the following:

• Adding a 'data' event handler to listen for data.
• Calling the stream.resume() method to explicitly open the flow.
• Calling the stream.pipe() method to send the data to a Writable.

You can switch back to paused mode by doing either of the following:

• If there are no pipe destinations, by calling the stream.pause() method.
• If there are pipe destinations, by removing any 'data' event handlers, and removing all pipe destinations by calling the stream.unpipe() method.

Note that, for backwards compatibility reasons, removing 'data' event handlers will not automatically pause the stream. Also, if there are piped destinations, then calling stream.pause() will not guarantee that the stream will remain paused once those destinations drain and ask for more data.

Examples of readable streams include:

• HTTP responses, on the client
• HTTP requests, on the server
• fs read streams
• zlib streams
• crypto streams
• TCP sockets
• child process stdout and stderr
• process.stdin

Event: 'close'#

Emitted when the stream and any of its underlying resources (a file descriptor, for example) have been closed. The event indicates that no more events will be emitted, and no further computation will occur.

Not all streams will emit the 'close' event as the 'close' event is optional.

Event: 'data'#

Attaching a 'data' event listener to a stream that has not been explicitly paused will switch the stream into flowing mode. Data will then be passed as soon as it is available.

If you just want to get all the data out of the stream as fast as possible, this is the best way to do so.

var readable = getReadableStreamSomehow();
readable.on('data', (chunk) => {
  console.log('got %d bytes of data', chunk.length);
});

Event: 'end'#

This event fires when there will be no more data to read.

Note that the 'end' event will not fire unless the data is completely consumed. This can be done by switching into flowing mode, or by calling stream.read() repeatedly until you get to the end.

var readable = getReadableStreamSomehow();
readable.on('data', (chunk) => {
  console.log('got %d bytes of data', chunk.length);
});
readable.on('end', () => {
  console.log('there will be no more data.');
});

Event: 'error'#

Emitted if there was an error receiving data.

Event: 'readable'#

When a chunk of data can be read from the stream, it will emit a 'readable' event.

In some cases, listening for a 'readable' event will cause some data to be read into the internal buffer from the underlying system, if it hadn't already.

var readable = getReadableStreamSomehow();
readable.on('readable', () => {
  // there is some data to read now
});

Once the internal buffer is drained, a 'readable' event will fire again when more data is available.

The 'readable' event is not emitted in the "flowing" mode with the sole exception of the last one, on end-of-stream.

The 'readable' event indicates that the stream has new information: either new data is available or the end of the stream has been reached. In the former case, stream.read() will return that data. In the latter case, stream.read() will return null. For instance, in the following example, foo.txt is an empty file:

const fs = require('fs');
var rr = fs.createReadStream('foo.txt');
rr.on('readable', () => {
  console.log('readable:', rr.read());
});
rr.on('end', () => {
  console.log('end');
});

The output of running this script is:

$ node test.js
readable: null
end

readable.isPaused()#

This method returns whether or not the readable has been explicitly paused by client code (using stream.pause() without a corresponding stream.resume()).

var readable = new stream.Readable

readable.isPaused() // === false
readable.pause()
readable.isPaused() // === true
readable.resume()
readable.isPaused() // === false

readable.pause()#

This method will cause a stream in flowing mode to stop emitting 'data' events, switching out of flowing mode. Any data that becomes available will remain in the internal buffer.

var readable = getReadableStreamSomehow();
readable.on('data', (chunk) => {
  console.log('got %d bytes of data', chunk.length);
  readable.pause();
  console.log('there will be no more data for 1 second');
  setTimeout(() => {
    console.log('now data will start flowing again');
    readable.resume();
  }, 1000);
});

readable.pipe(destination[, options])#

This method pulls all the data out of a readable stream, and writes it to the supplied destination, automatically managing the flow so that the destination is not overwhelmed by a fast readable stream.

Multiple destinations can be piped to safely.

var readable = getReadableStreamSomehow();
var writable = fs.createWriteStream('file.txt');
// All the data from readable goes into 'file.txt'
readable.pipe(writable);

This function returns the destination stream, so you can set up pipe chains like so:

var r = fs.createReadStream('file.txt');
var z = zlib.createGzip();
var w = fs.createWriteStream('file.txt.gz');
r.pipe(z).pipe(w);

For example, emulating the Unix cat command:

process.stdin.pipe(process.stdout);

By default stream.end() is called on the destination when the source stream emits 'end', so that destination is no longer writable. Pass { end: false } as options to keep the destination stream open.

This keeps writer open so that "Goodbye" can be written at the end.

reader.pipe(writer, { end: false });
reader.on('end', () => {
  writer.end('Goodbye\n');
});

Note that process.stderr and process.stdout are never closed until the process exits, regardless of the specified options.

readable.read([size])#

The read() method pulls some data out of the internal buffer and returns it. If there is no data available, then it will return null.

If you pass in a size argument, then it will return that many bytes. If size bytes are not available, then it will return null, unless we've ended, in which case it will return the data remaining in the buffer.

If you do not specify a size argument, then it will return all the data in the internal buffer.

This method should only be called in paused mode. In flowing mode, this method is called automatically until the internal buffer is drained.

var readable = getReadableStreamSomehow();
readable.on('readable', () => {
  var chunk;
  while (null !== (chunk = readable.read())) {
    console.log('got %d bytes of data', chunk.length);
  }
});

If this method returns a data chunk, then it will also trigger the emission of a 'data' event.

Note that calling stream.read([size]) after the 'end' event has been triggered will return null. No runtime error will be raised.

readable.resume()#

This method will cause the readable stream to resume emitting 'data' events.

This method will switch the stream into flowing mode. If you do not want to consume the data from a stream, but you do want to get to its 'end' event, you can call stream.resume() to open the flow of data.

var readable = getReadableStreamSomehow();
readable.resume();
readable.on('end', () => {
  console.log('got to the end, but did not read anything');
});

readable.setEncoding(encoding)#

Call this function to cause the stream to return strings of the specified encoding instead of Buffer objects. For example, if you do readable.setEncoding('utf8'), then the output data will be interpreted as UTF-8 data, and returned as strings. If you do readable.setEncoding('hex'), then the data will be encoded in hexadecimal string format.

This properly handles multi-byte characters that would otherwise be potentially mangled if you simply pulled the Buffers directly and called buf.toString(encoding) on them. If you want to read the data as strings, always use this method.

Also you can disable any encoding at all with readable.setEncoding(null). This approach is very useful if you deal with binary data or with large multi-byte strings spread out over multiple chunks.

var readable = getReadableStreamSomehow();
readable.setEncoding('utf8');
readable.on('data', (chunk) => {
  assert.equal(typeof chunk, 'string');
  console.log('got %d characters of string data', chunk.length);
});

readable.unpipe([destination])#

This method will remove the hooks set up for a previous stream.pipe() call.

If the destination is not specified, then all pipes are removed.

If the destination is specified, but no pipe is set up for it, then this is a no-op.

var readable = getReadableStreamSomehow();
var writable = fs.createWriteStream('file.txt');
// All the data from readable goes into 'file.txt',
// but only for the first second
readable.pipe(writable);
setTimeout(() => {
  console.log('stop writing to file.txt');
  readable.unpipe(writable);
  console.log('manually close the file stream');
  writable.end();
}, 1000);

readable.unshift(chunk)#

This is useful in certain cases where a stream is being consumed by a parser, which needs to "un-consume" some data that it has optimistically pulled out of the source, so that the stream can be passed on to some other party.

Note that stream.unshift(chunk) cannot be called after the 'end' event has been triggered; a runtime error will be raised.

If you find that you must often call stream.unshift(chunk) in your programs, consider implementing a Transform stream instead. (See API for Stream Implementors.)

// Pull off a header delimited by \n\n
// use unshift() if we get too much
// Call the callback with (error, header, stream)
const StringDecoder = require('string_decoder').StringDecoder;
function parseHeader(stream, callback) {
  stream.on('error', callback);
  stream.on('readable', onReadable);
  var decoder = new StringDecoder('utf8');
  var header = '';
  function onReadable() {
    var chunk;
    while (null !== (chunk = stream.read())) {
      var str = decoder.write(chunk);
      if (str.match(/\n\n/)) {
        // found the header boundary
        var split = str.split(/\n\n/);
        header += split.shift();
        var remaining = split.join('\n\n');
        var buf = Buffer.from(remaining, 'utf8');
        if (buf.length)
          stream.unshift(buf);
        stream.removeListener('error', callback);
        stream.removeListener('readable', onReadable);
        // now the body of the message can be read from the stream.
        callback(null, header, stream);
      } else {
        // still reading the header.
        header += str;
      }
    }
  }
}

Note that, unlike stream.push(chunk), stream.unshift(chunk) will not end the reading process by resetting the internal reading state of the stream. This can cause unexpected results if unshift() is called during a read (i.e. from within a stream._read() implementation on a custom stream). Following the call to unshift() with an immediate stream.push('') will reset the reading state appropriately, however it is best to simply avoid calling unshift() while in the process of performing a read.

readable.wrap(stream)#

Versions of Node.js prior to v0.10 had streams that did not implement the entire Streams API as it is today. (See Compatibility for more information.)

If you are using an older Node.js library that emits 'data' events and has a stream.pause() method that is advisory only, then you can use the wrap() method to create a Readable stream that uses the old stream as its data source.

You will very rarely ever need to call this function, but it exists as a convenience for interacting with old Node.js programs and libraries.

For example:

const OldReader = require('./old-api-module.js').OldReader;
const Readable = require('stream').Readable;
const oreader = new OldReader;
const myReader = new Readable().wrap(oreader);

myReader.on('readable', () => {
  myReader.read(); // etc.
});

Class: stream.Transform#

Transform streams are Duplex streams where the output is in some way computed from the input. They implement both the Readable and Writable interfaces.

Examples of Transform streams include:

• zlib streams
• crypto streams

Class: stream.Writable#

The Writable stream interface is an abstraction for a destination that you are writing data to.

Examples of writable streams include:

• HTTP requests, on the client
• HTTP responses, on the server
• fs write streams
• zlib streams
• crypto streams
• TCP sockets
• child process stdin
• process.stdout, process.stderr

Event: 'close'#

Emitted when the stream and any of its underlying resources (a file descriptor, for example) have been closed. The event indicates that no more events will be emitted, and no further computation will occur.

Not all streams will emit the 'close' event as the 'close' event is optional.

Event: 'drain'#

If a stream.write(chunk) call returns false, then the 'drain' event will indicate when it is appropriate to begin writing more data to the stream.

// Write the data to the supplied writable stream one million times.
// Be attentive to back-pressure.
function writeOneMillionTimes(writer, data, encoding, callback) {
  var i = 1000000;
  write();
  function write() {
    var ok = true;
    do {
      i -= 1;
      if (i === 0) {
        // last time!
        writer.write(data, encoding, callback);
      } else {
        // see if we should continue, or wait
        // don't pass the callback, because we're not done yet.
        ok = writer.write(data, encoding);
      }
    } while (i > 0 && ok);
    if (i > 0) {
      // had to stop early!
      // write some more once it drains
      writer.once('drain', write);
    }
  }
}

Event: 'error'#

Emitted if there was an error when writing or piping data.

Event: 'finish'#

When the stream.end() method has been called, and all data has been flushed to the underlying system, this event is emitted.

var writer = getWritableStreamSomehow();
for (var i = 0; i < 100; i ++) {
  writer.write('hello, #${i}!\n');
}
writer.end('this is the end\n');
writer.on('finish', () => {
  console.error('all writes are now complete.');
});

Event: 'pipe'#

This is emitted whenever the stream.pipe() method is called on a readable stream, adding this writable to its set of destinations.

var writer = getWritableStreamSomehow();
var reader = getReadableStreamSomehow();
writer.on('pipe', (src) => {
  console.error('something is piping into the writer');
  assert.equal(src, reader);
});
reader.pipe(writer);

Event: 'unpipe'#

This is emitted whenever the stream.unpipe() method is called on a readable stream, removing this writable from its set of destinations.

var writer = getWritableStreamSomehow();
var reader = getReadableStreamSomehow();
writer.on('unpipe', (src) => {
  console.error('something has stopped piping into the writer');
  assert.equal(src, reader);
});
reader.pipe(writer);
reader.unpipe(writer);

writable.cork()#

Forces buffering of all writes.

Buffered data will be flushed either at stream.uncork() or at stream.end() call.

writable.end([chunk][, encoding][, callback])#

Call this method when no more data will be written to the stream. If supplied, the callback is attached as a listener on the 'finish' event.

Calling stream.write() after calling stream.end() will raise an error.

// write 'hello, ' and then end with 'world!'
var file = fs.createWriteStream('example.txt');
file.write('hello, ');
file.end('world!');
// writing more now is not allowed!

writable.setDefaultEncoding(encoding)#

Sets the default encoding for a writable stream.

writable.uncork()#

Flush all data, buffered since stream.cork() call.

writable.write(chunk[, encoding][, callback])#

This method writes some data to the underlying system, and calls the supplied callback once the data has been fully handled. If an error occurs, the callback may or may not be called with the error as its first argument. To detect write errors, listen for the 'error' event.

The return value indicates if you should continue writing right now. If the data had to be buffered internally, then it will return false. Otherwise, it will return true.

This return value is strictly advisory. You MAY continue to write, even if it returns false. However, writes will be buffered in memory, so it is best not to do this excessively. Instead, wait for the 'drain' event before writing more data.

API for Stream Implementors#

To implement any sort of stream, the pattern is the same:

1. Extend the appropriate parent class in your own subclass via the extends keyword.
2. Call the appropriate parent class constructor in your constructor, to be sure that the internal mechanisms are set up properly.
3. Implement one or more specific methods, as detailed below.

The class to extend and the method(s) to implement depend on the sort of stream class you are writing:

Use-case	Class	Method(s) to implement
Reading only	Readable	_read
Writing only	Writable	_write, _writev
Reading and writing	Duplex	_read, _write, _writev
Operate on written data, then read the result	Transform	_transform, _flush

In your implementation code, it is very important to never call the methods described in API for Stream Consumers. Otherwise, you can potentially cause adverse side effects in programs that consume your streaming interfaces.

Class: stream.Duplex#

A "duplex" stream is one that is both Readable and Writable, such as a TCP socket connection.

Note that stream.Duplex is an abstract class designed to be extended with an underlying implementation of the stream._read(size) and stream._write(chunk, encoding, callback) methods as you would with a Readable or Writable stream class.

Since JavaScript doesn't have multiple prototypal inheritance, this class prototypally inherits from Readable, and then parasitically from Writable. It is thus up to the user to implement both the low-level stream._read(n) method as well as the low-level stream._write(chunk, encoding, callback) method on extension duplex classes.

new stream.Duplex(options)#

In classes that extend the Duplex class, make sure to call the constructor so that the buffering settings can be properly initialized.

Class: stream.PassThrough#

This is a trivial implementation of a Transform stream that simply passes the input bytes across to the output. Its purpose is mainly for examples and testing, but there are occasionally use cases where it can come in handy as a building block for novel sorts of streams.

Class: stream.Readable#

stream.Readable is an abstract class designed to be extended with an underlying implementation of the stream._read(size) method.

Please see API for Stream Consumers for how to consume streams in your programs. What follows is an explanation of how to implement Readable streams in your programs.

new stream.Readable([options])#

In classes that extend the Readable class, make sure to call the Readable constructor so that the buffering settings can be properly initialized.

readable._read(size)#

Note: Implement this method, but do NOT call it directly.

This method is prefixed with an underscore because it is internal to the class that defines it and should only be called by the internal Readable class methods. All Readable stream implementations must provide a _read method to fetch data from the underlying resource.

When _read() is called, if data is available from the resource, the _read() implementation should start pushing that data into the read queue by calling this.push(dataChunk). _read() should continue reading from the resource and pushing data until push returns false, at which point it should stop reading from the resource. Only when _read() is called again after it has stopped should it start reading more data from the resource and pushing that data onto the queue.

Note: once the _read() method is called, it will not be called again until the stream.push() method is called.

The size argument is advisory. Implementations where a "read" is a single call that returns data can use this to know how much data to fetch. Implementations where that is not relevant, such as TCP or TLS, may ignore this argument, and simply provide data whenever it becomes available. There is no need, for example to "wait" until size bytes are available before calling stream.push(chunk).

readable.push(chunk[, encoding])#

Note: This method should be called by Readable implementors, NOT by consumers of Readable streams.

If a value other than null is passed, The push() method adds a chunk of data into the queue for subsequent stream processors to consume. If null is passed, it signals the end of the stream (EOF), after which no more data can be written.

The data added with push() can be pulled out by calling the stream.read() method when the 'readable' event fires.

This API is designed to be as flexible as possible. For example, you may be wrapping a lower-level source which has some sort of pause/resume mechanism, and a data callback. In those cases, you could wrap the low-level source object by doing something like this:

// source is an object with readStop() and readStart() methods,
// and an `ondata` member that gets called when it has data, and
// an `onend` member that gets called when the data is over.


class SourceWrapper extends Readable {
  constructor(options) {
    super(options);

    this._source = getLowlevelSourceObject();

    // Every time there's data, we push it into the internal buffer.
    this._source.ondata = (chunk) => {
      // if push() returns false, then we need to stop reading from source
      if (!this.push(chunk))
        this._source.readStop();
    };

    // When the source ends, we push the EOF-signaling `null` chunk
    this._source.onend = () => {
      this.push(null);
    };
  }
  // _read will be called when the stream wants to pull more data in
  // the advisory size argument is ignored in this case.
  _read(size) {
    this._source.readStart();
  }
}

Example: A Counting Stream#

This is a basic example of a Readable stream. It emits the numerals from 1 to 1,000,000 in ascending order, and then ends.

const Readable = require('stream').Readable;

class Counter extends Readable {
  constructor(opt) {
    super(opt);
    this._max = 1000000;
    this._index = 1;
  }

  _read() {
    var i = this._index++;
    if (i > this._max)
      this.push(null);
    else {
      var str = '' + i;
      var buf = Buffer.from(str, 'ascii');
      this.push(buf);
    }
  }
}

Example: SimpleProtocol v1 (Sub-optimal)#

This is similar to the parseHeader function described here, but implemented as a custom stream. Also, note that this implementation does not convert the incoming data to a string.

However, this would be better implemented as a Transform stream. See SimpleProtocol v2 for a better implementation.

// A parser for a simple data protocol.
// The "header" is a JSON object, followed by 2 \n characters, and
// then a message body.
//
// NOTE: This can be done more simply as a Transform stream!
// Using Readable directly for this is sub-optimal. See the
// alternative example below under the Transform section.

const Readable = require('stream').Readable;

class SimpleProtocol extends Readable {
  constructor(source, options) {
    super(options);

    this._inBody = false;
    this._sawFirstCr = false;

    // source is a readable stream, such as a socket or file
    this._source = source;

    source.on('end', () => {
      this.push(null);
    });

    // give it a kick whenever the source is readable
    // read(0) will not consume any bytes
    source.on('readable', () => {
      this.read(0);
    });

    this._rawHeader = [];
    this.header = null;
  }

  _read(n) {
    if (!this._inBody) {
      var chunk = this._source.read();

      // if the source doesn't have data, we don't have data yet.
      if (chunk === null)
        return this.push('');

      // check if the chunk has a \n\n
      var split = -1;
      for (var i = 0; i < chunk.length; i++) {
        if (chunk[i] === 10) { // '\n'
          if (this._sawFirstCr) {
            split = i;
            break;
          } else {
            this._sawFirstCr = true;
          }
        } else {
          this._sawFirstCr = false;
        }
      }

      if (split === -1) {
        // still waiting for the \n\n
        // stash the chunk, and try again.
        this._rawHeader.push(chunk);
        this.push('');
      } else {
        this._inBody = true;
        var h = chunk.slice(0, split);
        this._rawHeader.push(h);
        var header = Buffer.concat(this._rawHeader).toString();
        try {
          this.header = JSON.parse(header);
        } catch (er) {
          this.emit('error', new Error('invalid simple protocol data'));
          return;
        }
        // now, because we got some extra data, unshift the rest
        // back into the read queue so that our consumer will see it.
        var b = chunk.slice(split);
        this.unshift(b);
        // calling unshift by itself does not reset the reading state
        // of the stream; since we're inside _read, doing an additional
        // push('') will reset the state appropriately.
        this.push('');

        // and let them know that we are done parsing the header.
        this.emit('header', this.header);
      }
    } else {
      // from there on, just provide the data to our consumer.
      // careful not to push(null), since that would indicate EOF.
      var chunk = this._source.read();
      if (chunk) this.push(chunk);
    }
  }
}
// Usage:
// var parser = new SimpleProtocol(source);
// Now parser is a readable stream that will emit 'header'
// with the parsed header data.

Class: stream.Transform#

A "transform" stream is a duplex stream where the output is causally connected in some way to the input, such as a zlib stream or a crypto stream.

There is no requirement that the output be the same size as the input, the same number of chunks, or arrive at the same time. For example, a Hash stream will only ever have a single chunk of output which is provided when the input is ended. A zlib stream will produce output that is either much smaller or much larger than its input.

Rather than implement the stream._read() and stream._write() methods, Transform classes must implement the stream._transform() method, and may optionally also implement the stream._flush() method. (See below.)

new stream.Transform([options])#

In classes that extend the Transform class, make sure to call the constructor so that the buffering settings can be properly initialized.

Events: 'finish' and 'end'#

The 'finish' and 'end' events are from the parent Writable and Readable classes respectively. The 'finish' event is fired after stream.end() is called and all chunks have been processed by stream._transform(), 'end' is fired after all data has been output which is after the callback in stream._flush() has been called.

transform._flush(callback)#

Note: This function MUST NOT be called directly. It MAY be implemented by child classes, and if so, will be called by the internal Transform class methods only.

In some cases, your transform operation may need to emit a bit more data at the end of the stream. For example, a Zlib compression stream will store up some internal state so that it can optimally compress the output. At the end, however, it needs to do the best it can with what is left, so that the data will be complete.

In those cases, you can implement a _flush() method, which will be called at the very end, after all the written data is consumed, but before emitting 'end' to signal the end of the readable side. Just like with stream._transform(), call transform.push(chunk) zero or more times, as appropriate, and call callback when the flush operation is complete.

This method is prefixed with an underscore because it is internal to the class that defines it, and should not be called directly by user programs. However, you are expected to override this method in your own extension classes.

transform._transform(chunk, encoding, callback)#

Note: This function MUST NOT be called directly. It should be implemented by child classes, and called by the internal Transform class methods only.

All Transform stream implementations must provide a _transform() method to accept input and produce output.

_transform() should do whatever has to be done in this specific Transform class, to handle the bytes being written, and pass them off to the readable portion of the interface. Do asynchronous I/O, process things, and so on.

Call transform.push(outputChunk) 0 or more times to generate output from this input chunk, depending on how much data you want to output as a result of this chunk.

Call the callback function only when the current chunk is completely consumed. Note that there may or may not be output as a result of any particular input chunk. If you supply a second argument to the callback it will be passed to the push method. In other words the following are equivalent:

transform.prototype._transform = function (data, encoding, callback) {
  this.push(data);
  callback();
};

transform.prototype._transform = function (data, encoding, callback) {
  callback(null, data);
};

This method is prefixed with an underscore because it is internal to the class that defines it, and should not be called directly by user programs. However, you are expected to override this method in your own extension classes.

Example: SimpleProtocol parser v2#

The example here of a simple protocol parser can be implemented simply by using the higher level Transform stream class, similar to the parseHeader and SimpleProtocol v1 examples.

In this example, rather than providing the input as an argument, it would be piped into the parser, which is a more idiomatic Node.js stream approach.

const Transform = require('stream').Transform;

class SimpleProtocol extends Transform {
  constructor(options) {
    super(options);

    this._inBody = false;
    this._sawFirstCr = false;
    this._rawHeader = [];
    this.header = null;
  }

  _transform(chunk, encoding, done) {
    if (!this._inBody) {
      // check if the chunk has a \n\n
      var split = -1;
      for (var i = 0; i < chunk.length; i++) {
        if (chunk[i] === 10) { // '\n'
          if (this._sawFirstCr) {
            split = i;
            break;
          } else {
            this._sawFirstCr = true;
          }
        } else {
          this._sawFirstCr = false;
        }
      }

      if (split === -1) {
        // still waiting for the \n\n
        // stash the chunk, and try again.
        this._rawHeader.push(chunk);
      } else {
        this._inBody = true;
        var h = chunk.slice(0, split);
        this._rawHeader.push(h);
        var header = Buffer.concat(this._rawHeader).toString();
        try {
          this.header = JSON.parse(header);
        } catch (er) {
          this.emit('error', new Error('invalid simple protocol data'));
          return;
        }
        // and let them know that we are done parsing the header.
        this.emit('header', this.header);

        // now, because we got some extra data, emit this first.
        this.push(chunk.slice(split));
      }
    } else {
      // from there on, just provide the data to our consumer as-is.
      this.push(chunk);
    }
    done();
  }
}
// Usage:
// var parser = new SimpleProtocol();
// source.pipe(parser)
// Now parser is a readable stream that will emit 'header'
// with the parsed header data.

Class: stream.Writable#

stream.Writable is an abstract class designed to be extended with an underlying implementation of the stream._write(chunk, encoding, callback) method.

Please see API for Stream Consumers for how to consume writable streams in your programs. What follows is an explanation of how to implement Writable streams in your programs.

new stream.Writable([options])#

In classes that extend the Writable class, make sure to call the constructor so that the buffering settings can be properly initialized.

writable._write(chunk, encoding, callback)#

All Writable stream implementations must provide a stream._write() method to send data to the underlying resource.

Note: This function MUST NOT be called directly. It should be implemented by child classes, and called by the internal Writable class methods only.

Call the callback using the standard callback(error) pattern to signal that the write completed successfully or with an error.

If the decodeStrings flag is set in the constructor options, then chunk may be a string rather than a Buffer, and encoding will indicate the sort of string that it is. This is to support implementations that have an optimized handling for certain string data encodings. If you do not explicitly set the decodeStrings option to false, then you can safely ignore the encoding argument, and assume that chunk will always be a Buffer.

This method is prefixed with an underscore because it is internal to the class that defines it, and should not be called directly by user programs. However, you are expected to override this method in your own extension classes.

writable._writev(chunks, callback)#

Note: This function MUST NOT be called directly. It may be implemented by child classes, and called by the internal Writable class methods only.

This function is completely optional to implement. In most cases it is unnecessary. If implemented, it will be called with all the chunks that are buffered in the write queue.

Simplified Constructor API#

In simple cases there is now the added benefit of being able to construct a stream without inheritance.

This can be done by passing the appropriate methods as constructor options:

Examples:

Duplex#

var duplex = new stream.Duplex({
  read: function(n) {
    // sets this._read under the hood

    // push data onto the read queue, passing null
    // will signal the end of the stream (EOF)
    this.push(chunk);
  },
  write: function(chunk, encoding, next) {
    // sets this._write under the hood

    // An optional error can be passed as the first argument
    next()
  }
});

// or

var duplex = new stream.Duplex({
  read: function(n) {
    // sets this._read under the hood

    // push data onto the read queue, passing null
    // will signal the end of the stream (EOF)
    this.push(chunk);
  },
  writev: function(chunks, next) {
    // sets this._writev under the hood

    // An optional error can be passed as the first argument
    next()
  }
});

Readable#

var readable = new stream.Readable({
  read: function(n) {
    // sets this._read under the hood

    // push data onto the read queue, passing null
    // will signal the end of the stream (EOF)
    this.push(chunk);
  }
});

Transform#

var transform = new stream.Transform({
  transform: function(chunk, encoding, next) {
    // sets this._transform under the hood

    // generate output as many times as needed
    // this.push(chunk);

    // call when the current chunk is consumed
    next();
  },
  flush: function(done) {
    // sets this._flush under the hood

    // generate output as many times as needed
    // this.push(chunk);

    done();
  }
});

Writable#

var writable = new stream.Writable({
  write: function(chunk, encoding, next) {
    // sets this._write under the hood

    // An optional error can be passed as the first argument
    next()
  }
});

// or

var writable = new stream.Writable({
  writev: function(chunks, next) {
    // sets this._writev under the hood

    // An optional error can be passed as the first argument
    next()
  }
});

Streams: Under the Hood#

Buffering#

Both Writable and Readable streams will buffer data on an internal object which can be retrieved from _writableState.getBuffer() or _readableState.buffer, respectively.

The amount of data that will potentially be buffered depends on the highWaterMark option which is passed into the constructor.

Buffering in Readable streams happens when the implementation calls stream.push(chunk). If the consumer of the Stream does not call stream.read(), then the data will sit in the internal queue until it is consumed.

Buffering in Writable streams happens when the user calls stream.write(chunk) repeatedly, even when it returns false.

The purpose of streams, especially with the stream.pipe() method, is to limit the buffering of data to acceptable levels, so that sources and destinations of varying speed will not overwhelm the available memory.

Compatibility with Older Node.js Versions#

In versions of Node.js prior to v0.10, the Readable stream interface was simpler, but also less powerful and less useful.

• Rather than waiting for you to call the stream.read() method, 'data' events would start emitting immediately. If you needed to do some I/O to decide how to handle data, then you had to store the chunks in some kind of buffer so that they would not be lost.
• The stream.pause() method was advisory, rather than guaranteed. This meant that you still had to be prepared to receive 'data' events even when the stream was in a paused state.

In Node.js v0.10, the Readable class was added. For backwards compatibility with older Node.js programs, Readable streams switch into "flowing mode" when a 'data' event handler is added, or when the stream.resume() method is called. The effect is that, even if you are not using the new stream.read() method and 'readable' event, you no longer have to worry about losing 'data' chunks.

Most programs will continue to function normally. However, this introduces an edge case in the following conditions:

• No 'data' event handler is added.
• The stream.resume() method is never called.
• The stream is not piped to any writable destination.

For example, consider the following code:

// WARNING!  BROKEN!
net.createServer((socket) => {

  // we add an 'end' method, but never consume the data
  socket.on('end', () => {
    // It will never get here.
    socket.end('I got your message (but didnt read it)\n');
  });

}).listen(1337);

In versions of Node.js prior to v0.10, the incoming message data would be simply discarded. However, in Node.js v0.10 and beyond, the socket will remain paused forever.

The workaround in this situation is to call the stream.resume() method to start the flow of data:

// Workaround
net.createServer((socket) => {

  socket.on('end', () => {
    socket.end('I got your message (but didnt read it)\n');
  });

  // start the flow of data, discarding it.
  socket.resume();

}).listen(1337);

In addition to new Readable streams switching into flowing mode, pre-v0.10 style streams can be wrapped in a Readable class using the stream.wrap() method.

Object Mode#

Normally, Streams operate on Strings and Buffers exclusively.

Streams that are in object mode can emit generic JavaScript values other than Buffers and Strings.

A Readable stream in object mode will always return a single item from a call to stream.read(size), regardless of what the size argument is.

A Writable stream in object mode will always ignore the encoding argument to stream.write(data, encoding).

The special value null still retains its special value for object mode streams. That is, for object mode readable streams, null as a return value from stream.read() indicates that there is no more data, and stream.push(null) will signal the end of stream data (EOF).

No streams in Node.js core are object mode streams. This pattern is only used by userland streaming libraries.

You should set objectMode in your stream child class constructor on the options object. Setting objectMode mid-stream is not safe.

For Duplex streams objectMode can be set exclusively for readable or writable side with readableObjectMode and writableObjectMode respectively. These options can be used to implement parsers and serializers with Transform streams.

const StringDecoder = require('string_decoder').StringDecoder;
const Transform = require('stream').Transform;

// Gets \n-delimited JSON string data, and emits the parsed objects
class JSONParseStream extends Transform {
  constructor() {
    super({ readableObjectMode : true });

    this._buffer = '';
    this._decoder = new StringDecoder('utf8');
  }

  _transform(chunk, encoding, cb) {
    this._buffer += this._decoder.write(chunk);
    // split on newlines
    var lines = this._buffer.split(/\r?\n/);
    // keep the last partial line buffered
    this._buffer = lines.pop();
    for (var l = 0; l < lines.length; l++) {
      var line = lines[l];
      try {
        var obj = JSON.parse(line);
      } catch (er) {
        this.emit('error', er);
        return;
      }
      // push the parsed object out to the readable consumer
      this.push(obj);
    }
    cb();
  }

  _flush(cb) {
    // Just handle any leftover
    var rem = this._buffer.trim();
    if (rem) {
      try {
        var obj = JSON.parse(rem);
      } catch (er) {
        this.emit('error', er);
        return;
      }
      // push the parsed object out to the readable consumer
      this.push(obj);
    }
    cb();
  }
}

stream.read(0)#

There are some cases where you want to trigger a refresh of the underlying readable stream mechanisms, without actually consuming any data. In that case, you can call stream.read(0), which will always return null.

If the internal read buffer is below the highWaterMark, and the stream is not currently reading, then calling stream.read(0) will trigger a low-level stream._read() call.

There is almost never a need to do this. However, you will see some cases in Node.js's internals where this is done, particularly in the Readable stream class internals.

stream.push('')#

Pushing a zero-byte string or Buffer (when not in Object mode) has an interesting side effect. Because it is a call to stream.push(), it will end the reading process. However, it does not add any data to the readable buffer, so there's nothing for a user to consume.

Very rarely, there are cases where you have no data to provide now, but the consumer of your stream (or, perhaps, another bit of your own code) will know when to check again, by calling stream.read(0). In those cases, you may call stream.push('').

So far, the only use case for this functionality is in the tls.CryptoStream class, which is deprecated in Node.js/io.js v1.0. If you find that you have to use stream.push(''), please consider another approach, because it almost certainly indicates that something is horribly wrong.