Node js, piping pdfkit to a memory stream

An updated answer for 2020. There is no need to introduce a new memory stream because "PDFDocument instances are readable Node streams".

You can use the get-stream package to make it easy to wait for the document to finish before passing the result back to your caller. https://www.npmjs.com/package/get-stream

const PDFDocument = require('pdfkit')
const getStream = require('get-stream')

const pdf = () => {
  const doc = new PDFDocument()
  doc.text('Hello, World!')
  doc.end()
  return await getStream.buffer(doc)
}


// Caller could do this:
const pdfBuffer = await pdf()
const pdfBase64string = pdfBuffer.toString('base64')

You don't have to return a buffer if your needs are different. The get-stream readme offers other examples.

There's no need to use an intermediate memory stream¹ – just pipe the pdfkit output stream directly into a HTTP upload stream.

In my experience, the AWS SDK is garbage when it comes to working with streams, so I usually use request.

var upload = request({
    method: 'PUT',
    url: 'https://bucket.s3.amazonaws.com/doc.pdf',
    aws: { bucket: 'bucket', key: ..., secret: ... }
});

doc.pipe(upload);

^{1 - in fact, it is usually undesirable to use a memory stream because that means buffering the entire thing in RAM, which is exactly what streams are supposed to avoid!}

You could try something like this, and upload it to S3 inside the end event.

var doc = new pdfkit();

var MemoryStream = require('memorystream');
var memStream = new MemoryStream(null, {
   readable : false
});

doc.pipe(memStream);

doc.on('end', function () {
   var buffer = Buffer.concat(memStream.queue);
   awsservice.putS3Object(buffer, fileName, fileType, folder).then(function () { }, reject);
})