Node.js Using Streams Why Streams?


Lets examine the following two examples for reading a file's contents:

The first one, which uses an async method for reading a file, and providing a callback function which is called once the file is fully read into the memory:

fs.readFile(`${__dirname}/utils.js`, (err, data) => {
  if (err) {
  } else {

And the second, which uses streams in order to read the file's content, piece by piece:

var fileStream = fs.createReadStream(`${__dirname}/file`);
var fileContent = '';
fileStream.on('data', data => {
  fileContent += data.toString();

fileStream.on('end', () => {

fileStream.on('error', err => {

It's worth mentioning that both examples do the exact same thing. What's the difference then?

  • The first one is shorter and looks more elegant
  • The second lets you do some processing on the file while it is being read (!)

When the files you deal with are small then there is no real effect when using streams, but what happens when the file is big? (so big that it takes 10 seconds to read it into memory)

Without streams you'll be waiting, doing absolutely nothing (unless your process does other stuff), until the 10 seconds pass and the file is fully read, and only then you can start processing the file.

With streams, you get the file's contents piece by piece, right when they're available - and that lets you process the file while it is being read.

The above example does not illustrate how streams can be utilized for work that cannot be done when going the callback fashion, so lets look at another example:

I would like to download a gzip file, unzip it and save its content to the disk. Given the file's url this is what's need to be done:

  • Download the file
  • Unzip the file
  • Save it to disk

Here's a [small file][1], which is stored in my S3 storage. The following code does the above in the callback fashion.

var startTime =
s3.getObject({Bucket: 'some-bucket', Key: 'tweets.gz'}, (err, data) => {
  // here, the whole file was downloaded

  zlib.gunzip(data.Body, (err, data) => {
    // here, the whole file was unzipped

    fs.writeFile(`${__dirname}/tweets.json`, data, err => {
      if (err) console.error(err)

      // here, the whole file was written to disk
      var endTime =
      console.log(`${endTime - startTime} milliseconds`) // 1339 milliseconds

// 1339 milliseconds

This is how it looks using streams:

s3.getObject({Bucket: 'some-bucket', Key: 'tweets.gz'}).createReadStream()

// 1204 milliseconds

Yep, it's not faster when dealing with small files - the tested file weights 80KB. Testing this on a bigger file, 71MB gzipped (382MB unzipped), shows that the streams version is much faster

  • It took 20925 milliseconds to download 71MB, unzip it and then write 382MB to disk - using the callback fashion.
  • In comparison, it took 13434 milliseconds to do the same when using the streams version (35% faster, for a not-so-big file)