Making PDF’s with and without XSL FO (Revised)

date:2006-03-20 15:41:39 category:DocBook

I’ve been using the Apache FOP [link ] to transform XSL:FO into PDF’s. FOP works well enough for some DocBook processing, but it can’t handle really complex stuff gracefully.

Can we also use this XSL:FO technology for producing PDF form letters as part of a Customer Relationship Management solution? Is FOP suitable?

XSL:FO Products

A very complete list is available at XMLSoftware.com [link ].

The core list of tools include the following:

• Adobe seems to own this space; they offer XSL:FO processing tools in a wide variety of forms. Further, they have an on-line service [link ] that can convert many things into PDF’s.
• ReportLab [link ] has an engine that will create PDF’s from source files. They also have an open source toolkit.
• Antenna House [link ] offers the XSL Formatter product that seems to be widely regarded.
• RenderX [link ] offers the XEP product that does this job well.
• Chive [link ] has their Apoc product, which is a .Net solution.
• Ecrion [link ] also has a .NET product.
• Alt-soft [link ] has a beta product for xml2pdf conversion.
• Stylus Studio [link ] has a tool that edits XSL:FO and runs Apache FOP for you.
• The W3Schools site lists two more tools [link ]: Xinc [link ] and Scriptura [link ]. Lunasil looks like a very small company; Xinc appears to be their only product. Scriptura’s description sounds like Adobe’s – it’s tough to go head-to-head with the leader.

A nice comparison between the market-leaders XEP and XSL Formatter is available at XSLT.com [link ].

Beyond these tools, Altova has a product [link ] to help design the XSL that controls document production.

Finally, a partial solution is iText [link ], which is a set of Java libraries that create PDF’s.

Personally, I would prefer to use something like ReportLab Toolkit [link ], but the customer will likely balk at anything non-Java. Indeed, between Python, Cheetah and ReportLab, I think I’d be done with a solution, rather than still studying the problem.

Additional Thoughts. The real question, however, is why XSL:FO? There are lots of products that produce PDF files. We have to look at the real purpose behind the task. For converting DocBook files, the stylesheets and XSL:FO components fit into a very nice processing pipeline that is driven by Xalan.

However, in the case of Customer Relationship Management, we are using an application to do “mail-merge” processing, combining a template letter with details extracted from the database. We can imagine two families of architectures for this:

Minimal Markup. In this case, either no markup or a home-brewed minimal markup language are used. This has the advantage of minimizing knowledge of XSL:FO.

1. Custom software, using iText or similar, that has the letter template encoded in the application. This is simple, but quite inflexible.
2. Custom software, using iText or similar, that has the letter template in some home-brewed markup language. It reads the template and the data, merges the data in and produces PDF output. In this case, much of the formatting information is either implied or encoded in the home-brewed markup language.

Existing Markup. In this case, existing markup languages are leveraged.

1. Template processing. Use an existing template engine (Velocity, Cheetah or similar) that uses a letter template in a meta-markup language. This produces XSL:FO output that can then be processed by any of the XSL:FO tools.
2. XSL processing. Use Xalan to transform the source details into individual letters using an XSL template for the letter. This produces XSL:FO output that can then be processed by any of the XSL:FO tools.

Adobe has products that handle all of this, but they can be expensive. The Minimal Markup solutions involve some programming, but it is the kind of inflexible programming that makes every minor change to the template a huge cost and expense. The Existing Markup solutions involve considerable learning, but also mean that letter changes are handled by someone who knows the template language and the XSL:FO markup language.

The knowledge of XSL:FO can be minimized with clever use of the Cheetah or Velocity templates.