Files downloaded from Amazon S3 using Knox and Node.js are corrupt

I'm using knox to access my Amazon S3 bucket for file storage. I'm storing all kinds of files - mostly MS Office and pdfs but could be binary or any other kind. I'm also using express 4.13.3 and busboy with connect-busboy for streaming support; when uploading file I'm handling with busboy and thence direct to S3 via knox, so avoiding having to write them to local disk first.

The files upload fine (I can browse and download them manually using Transmit) but I'm having problems downloading.

For clarity I don't want to write the file to local disk, instead keeping it in an in-memory buffer. Here's the code I'm using to handle the GET request:

// instantiate a knox object
var s3client = knox.createClient({
  key: config.AWS.knox.key,
  secret: config.AWS.knox.secret,
  bucket: config.AWS.knox.bucket,
  region: config.AWS.region
});

var buffer = undefined;

s3client.get(path+'/'+fileName)
.on('response', function(s3res){

  s3res.setEncoding('binary');

  s3res.on('data', function(chunk){
    buffer += chunk;
  });

  s3res.on('end', function() {
    buffer = new Buffer(buffer, 'binary');
    var fileLength = buffer.length;
    res.attachment(fileName);
    res.append('Set-Cookie', 'fileDownload=true; path=/');
    res.append('Content-Length', fileLength);
    res.status(s3res.statusCode).send(buffer);
  });

}).end();

The file downloads to the browser - I'm using John Culviner's jquery.fileDownload.js - but what is downloaded is corrupt and can't be opened. As you can see I'm using express' .attachment to set the headers for mime type and .append for the additional headers (using .set instead makes no difference).

When the file downloads in Chrome I see the message 'Resource interpreted as Document but transferred with MIME type application/vnd.openxmlformats-officedocument.spreadsheetml.sheet:' (for an Excel file), so express is setting the header correctly, and the size of the file downloaded matches that I see when examining the bucket.

Any ideas what's going wrong?

Looks like the contents might not be being sent to the browser as binary. Try something like the following:

if (s3Res.headers['content-type']) {
  res.type( s3Res.headers['content-type'] );
}
res.attachment(fileName);

s3Res.setEncoding('binary');
s3Res.on('data', function(data){
  res.write(data, 'binary');
});

s3Res.on('end', function() {
  res.send();
});

It will also send the data one chunk at a time as it comes in, so it should be a bit more memory efficient.