Serve dynamically generated PDF with remote images in Node.js

Tyler Eich picture Tyler Eich · Apr 6, 2014 · Viewed 15.1k times · Source

I'm trying to create a Node server that generates a PDF on-the-fly using PDFKit. The PDF is generated based on parameters from a POST request (via Express). One of the parameters specifies an image URL, which the server downloads and injects into the PDF.

Right now, I have the following structure:

// Get dependencies
var express = require('express'),
http = require('http'),
fs = require('fs'),
pdfDocument = require('pdfkit');

// Get express started.
var app = express();

// Use JSON in POST body
app.use(express.json());

// Setup POST response
app.post('/post_pdf', function(req, res) {
    // Get the PDF initialized
    var doc = new pdfDocument();

    // Set some headers
    res.statusCode = 200;
    res.setHeader('Content-type', 'application/pdf');
    res.setHeader('Access-Control-Allow-Origin', '*');

    // Header to force download
    res.setHeader('Content-disposition', 'attachment; filename=Untitled.pdf');     

    // Pipe generated PDF into response
    doc.pipe(res);

    /**
     * Generate PDF contents
     */

    // Prepare write stream for image
    var image = fs.createWriteStream('image.jpeg');

    // Download image
    http.get("http://dummyimage.com/640.jpeg", function(response) {

        // Pipe response into image write stream
        // (because PDFKit needs to read from a saved file)
        response.pipe(image).on('close', function() {

            // Read data back, make sure there are no errors
            fs.readFile('image.jpeg', function(err, data) {
                if (err) throw err;

                /**
                 * Use `data` to get image info (width, height, etc.)
                 * ------------------
                 * Inject image
                 */

                // Close document and response
                doc.end();
                res.end();
                return;
            })
        });
    });
});

I have two questions:

  • Is there a less messy way to do this, perhaps with fewer nested callbacks? I'm totally open to adding another dependency to make life easier.
  • Right now, the code above does not work. It returns a PDF, but the PDF is corrupted (according to Preview). Any tips as to why this could be occurring are very welcome.

Answer

Tyler Eich picture Tyler Eich · Apr 7, 2014

In debugging this issue, I discovered several things:

PDFKit does not need to read info from a file. It will also accept a Buffer

doc.image(myBuffer); // You don't have to use a path string

When piping a file directly into the response, a manual call to response.end() will cause problems if the file has already been closed

doc.pipe(res); // Pipe document directly into the response

doc.end(); // When called, this ends the file and the response

// res.end(); <-- DON'T call res.end()
//                The response was already closed by doc.end()
return;

Request is a super-useful NodeJS library that can flatten the callback tree


Updated code:

var express = require('express'),
request = require('request'),
pdfDocument = require('pdfkit');

// Start Express
var app = express();

// Use JSON in POST body
app.use(express.json());

// Setup POST response
app.post('/post_pdf', function(req, res) {
    // Create PDF
    var doc = new pdfDocument();

    // Write headers
    res.writeHead(200, {
        'Content-Type': 'application/pdf',
        'Access-Control-Allow-Origin': '*',
        'Content-Disposition': 'attachment; filename=Untitled.pdf'
    });

    // Pipe generated PDF into response
    doc.pipe(res);

    // Process image
    request({
        url: 'http://dummyimage.com/640.jpeg',
        encoding: null // Prevents Request from converting response to string
    }, function(err, response, body) {
        if (err) throw err;

        // Inject image
        doc.image(body); // `body` is a Buffer because we told Request
                         // to not make it a string

        doc.end(); // Close document and, by extension, response
        return;
    });
});