Published on ONJava.com (http://www.onjava.com/)
 See this if you're having trouble printing code examples

Dynamically Creating PDFs in a Web Application

by Sean C. Sullivan

On a recent logistics project, a customer asked our team to build a web site that would allow users to query a legacy system for shipment information. The customer defined three main requirements:

  1. The shipping information had to be returned in the form of a PDF document.
  2. The PDF file must be downloadable through a web browser.
  3. The PDF file must be viewable using the Adobe Acrobat Reader.

Our team had plenty of experience with J2EE web applications, but we had little experience with PDF documents. We needed to find a pure Java class library that could produce sophisticated PDF documents in a server-side web application. We found a solution that completely met our needs: iText.

iText Class Library

iText is an open source pure Java class library for creating and manipulating PDF documents. Bruno Lowagie and Paulo Soares lead the project. The iText API enables a Java developer to programmatically create PDF documents. iText delivers a rich set of features:

Related Reading

Java Servlet Programming
By Jason Hunter

iText is an open source library. At the time of this writing, the iText software is available under a dual license: the Mozilla Public License (MPL) and the LGPL. Consult the iText web site for details. In this article, you'll see the iText API in action. We will demonstrate how to use iText and servlets to dynamically generate PDF documents in a server-side application.

Getting Started

First, you will need to obtain the iText JAR file. Visit the iText web site and download the current release. At the time of this writing, the current iText release is version 0.99. The iText web site provides API documentation and a comprehensive tutorial.

In addition to iText, we'll be using servlets, too. If you aren't familiar with servlets, you can learn about them in Jason Hunter's book, Java Servlet Programming. You will need to obtain a J2EE application server or a standalone servlet engine. Some good open source options are Tomcat, Jetty, and JBoss. The rest of this article assumes that you are using Jakarta Tomcat 4.1.

The iText API

The iText API is intuitive and easy to use. Using iText, you will be able to programmatically create customized PDF documents. The iText library consists of the following packages:


For generating PDF files, you'll need only com.lowagie.text and com.lowagie.text.pdf.

Our example application uses these iText classes:


The key classes are Document and PdfWriter. You will always use both of these classes when creating PDF documents. Document is an object-oriented representation of a PDF document. You can add content to the document by invoking methods provided by the Document class. A PdfWriter object associates a Document with a java.io.OutputStream object.

Coordinate System for iText Documents

When I wrote my first iText program, I stumbled over the coordinate system. I naively assumed that iText's coordinate system was identical to Swing's coordinate system. This is not the case.

In Swing, the origin (0, 0) is located in the upper left-hand corner of a component. In iText, the origin is located in the bottom left-hand corner of a page.

Using iText in a Web Application

During your design phase, you must decide how you plan to use iText. I've built web applications using both of the following techniques.

Technique A

Create the PDF file on the server's filesystem. The application uses java.io.FileOutputStream to write the file to the server's filesystem. The user will download the file via HTTP GET.

Technique B

Create the PDF file in memory using java.io.ByteArrayOutputStream. The application sends the PDF bytes to the client via the servlet's output stream.

Source Code

Download the source code for this example:

I prefer technique B to technique A because the application does not write to the server's filesystem, and the application is guaranteed to work in a clustered server environment. Technique A can fail if your application runs in a clustered environment, and the server cluster does not provide session affinity.

Example: PDFServlet

Our example application consists of a single class: PDFServlet. This servlet uses technique B from the previous section. The OutputStream is a java.io.ByteArrayOutputStream. With ByteArrayOutputStream, the PDF document bytes will be in memory. When PDFServlet receives an HTTP request, it will dynamically generate a PDF document and send the document to the client.

The PDFServlet class extends javax.servlet.http.HttpServlet and imports two of the iText packages, com.lowagie.text and com.lowagie.text.pdf.

The doGet Method

Most servlets override either the doPost method or the doGet method. Our servlet is no different. The PDFServlet class overrides the doGet method. The servlet will generate a PDF file any time it receives an incoming HTTP GET request.

In a nutshell, the servlet's doGet method does the following:

  1. Creates a ByteArrayOutputStream object that contains the PDF document bytes.
  2. Sets the HTTP response headers on the response object.
  3. Gets the servlet output stream.
  4. Writes the document bytes to the servlet output stream.
  5. Flushes the servlet output stream.

Figure 1. Editing doGet in Eclipse

The generatePDFDocumentBytes Method

The generatePDFDocumentBytes method is responsible for creating the PDF document. The three most important objects in this method are the Document object, the ByteArrayOutputStream object, and the PdfWriter object. The PdfWriter associates the Document with the ByteArrayOutputStream.

Document doc = new Document();
ByteArrayOutputStream baosPDF = new ByteArrayOutputStream();
PdfWriter docWriter = null;
docWriter = PdfWriter.getInstance(doc, baosPDF);
// ...

Adding content to a Document is done with the add method.

doc.add(new Paragraph(
    "This document was created by a class named: "
    + this.getClass().getName()));

doc.add(new Paragraph(
    "This document was created on "
    + new java.util.Date()));

When you are done adding content, close the Document and PdfWriter objects.


After closing the document, the ByteArrayOutputStream object is returned to the caller.

return baosPDF;

The ByteArrayOutputStream contains all bytes for the PDF document.

HTTP Response Headers

In this application, we care only about four HTTP response headers: Content-type, Content-disposition, Content-length, and Cache-control. If you've never worked with HTTP headers before, consult the HTTP 1.1 specification.

Examine the doGet method in the PDFServlet. You'll notice that the HTTP response headers are set before any data is written to the servlet output stream. This is an important, yet subtle, point.

Let's look at each response header in more detail.


In servlets, HttpServletResponse has a content type that indicates the type of content that the response contains. For PDF files, the content type is application/pdf. If the servlet does not set a content type, the web browser may have a difficult time determining how to handle the file.

PDFServlet sets the content type with the following line:



The Content-disposition header provides information that helps a web browser identify the content of the HTTP response. When a web browser reads this header, it can determine:

RFC 2183 provides a full explanation of the Content-disposition header.

By setting the Content-disposition header appropriately, the servlet can instruct the browser to display the file "inline," or to treat it like an attachment.

Example 1. Displaying a file inline

Content-disposition: inline; filename=foobar.pdf

Example 2. Attaching a file to the response

Content-disposition: attachment; filename=foobar.pdf

The following pseudo-code demonstrates how to set the header:

public void doGet(HttpServletRequest req, HttpServletResponse resp)
	// ... 
		"inline; filename=foobar.pdf" );
	// ... 

Cache-Control Headers

Depending upon the nature of your application, you may or may not want web browsers to cache the PDF files that you are generating. There are a variety of HTTP headers that a server-side web application can use to control caching of content. Some examples are:

A full explanation of Cache-Control headers is found in the HTTP 1.1 specification.

The PDFServlet sets Cache-Control to max-age=30. This header tells the web browser to cache the file for a maximum of 30 seconds.


The Content-length header must be set to the number of bytes in the PDF file. If the Content-length header is not set correctly, the web browser may not be able to display the file. Example code might be:

ByteArrayOutputStream baos = getByteArrayOutputStream();

Sending the PDF Document to a Web Browser

PDFServlet sends the PDF document to the client by writing bytes to the servlet's output stream. It obtains the output stream by calling getOutputStream() on the HttpServletResponse object. getOutputStream returns an object of type javax.servlet.ServletOutputStream.

ServletOutputStream sos;
sos = resp.getOutputStream();

After writing all data to the stream, call the flush() method to send all bytes to the client.

Packaging and Deployment

To run the PDFServlet in Tomcat, you'll need to package the application in a WAR file. The iText JAR file (itext-0.99.jar) must be placed in the WAR file's lib directory. If you forget to include the iText JAR file, the servlet will fail with a java.lang.NoClassDefFoundError.

Running the Application

After the WAR file has been deployed, you are ready to test the servlet. Jakarta Tomcat listens for requests on port 8080.

Point your web browser to http://hostname:8080/pdfservlet/createpdf.

When you visit the URL, the servlet executes and sends a PDF document back to your browser.

Beyond iText

iText provides a great low-level API for producing PDF documents. However, it may not be the best tool for every application.

At my day job, we used iText in combination with Microsoft Word and Adobe Acrobat. First, our team designed a shipment form using Microsoft Word. Next, we converted the Word document to PDF using Adobe Acrobat. Then, using iText's template capability, we loaded the PDF file into our application. From there, it was quite easy to fill in data values on the form and output the final PDF document.

For report-oriented web applications, tools like JasperReports provide a higher level of abstraction than iText.


When your Java application needs to dynamically create PDF documents, the iText class library is a great solution. You can experiment with iText's capabilities by enhancing and extending the code in this article. In a short time, you'll be able to impress your co-workers and customers with sophisticated PDF documents.

Additional Resources

If you are exploring Microsoft's .NET platform, be sure to check out iTextdotNet and iTextSharp. Both projects are derived from the Java-based iText library. iTextSharp is written in Microsoft's C# language.

Sean C. Sullivan has been developing Internet applications with Java since 1996. His recent work includes B2B web applications, various open source projects, and the development of an Internet e-commerce payment system at Intel.

Return to ONJava.com.

Copyright © 2009 O'Reilly Media, Inc.