Say No To Powerpoint
This isn't something that Aunt Tilly would do, but I already know XHTML and a bit of CSS, so authoring becomes very, very fast. I dash off a few h1, ul, and img, and I am done.
Your Mission, Should You Choose to Accept It
Next semester, I will useUbiquitous Presenter for “active learning” in the classroom. It is a nifty system where both the professor and the students can annotate slides. However, the slides must be in Powerpoint format (and, I am told, it is best if the Powerpoint is not too old or too new or “too complex”). As an alternative, the system will grudgingly accept a set of 576x432 pixel image files. (It took me a long time to realize that this means 8x6 inches at 72 dots per inch.)
I forged ahead with this recipe:
- Use PrinceXML to convert XHTML to PDF
- Use Pdftk to burst the PDF into a sequence of single-page PDFs
- Use ImageMagick to convert the pages to PNG
It worked, but the image quality was rather poor.
There had to be a better way. My mission was to find one.
I considered using JEditorPane, but I had not been happy with it in the past. It has a lot of quirks and doesn't seem very actively maintained. Then I happened to spot Flying Saucer.
Flying Saucer is “an XML/XHTML/CSS 2.1 renderer in 100% Java”. It renders to Java2D images, and, together with the iText library, to PDF. (It does not try to process arbitrary HTML, so you won't want to build a web browser with it.) Some people use it to produce PDF or images on the fly, for example in web applications. Generate XML or XHTML, use CSS for styling, and you can produce quite a variety of output with very little work.
I loaded a slide sample into the Flying Saucer viewer application. Impressively enough, all of the CSS was correctly interpreted. Make that “almost all of the CSS”. A bug with image scaling was a showstopper, but I sent a message to the mailing list, and a patch materialized within hours.
Flying Saucer comes with command-line utilities to generate PDF or images, but there was nothing to produce an image per page. In a day, I hacked together a simple program for this purpose, by copying the pagination code from the PDF generator and the image rendering code from the image generator. I don't understand all the mumbo-jumbo with devices and contexts, but it works perfectly. Here is a slide sample:
If you ever need to render images or PDF with lots of text and a fairly rigid structure, give Flying Saucer a try.
What does this all mean in a bigger way? It is an issue of infrastructure. The Ubiquitous Presenter tools are tethered to Powerpoint. It is a fragile infrastructure. There is no reliable mechanism to process Powerpoint other than Powerpoint, and even that comes with the proviso that the files are neither too old nor too new.
In contrast, open standards and open source enabled me to leverage a huge infrastructure. When I submit a bug report about CSS, there was no arguing about the expected behavior, but the bug simply got fixed. (Try that with Powerpoint!) There are lots of tools that produce or consume XML+CSS. The Flying Saucer people integrate the iText stuff; I take the Flying Saucer stuff and solve my problem in a day. It's all open source and there is very little friction. It isn't something that I think about very much, just like I don't usually spend a lot of time thinking about the physical infrastructure that surrounds us. As long as it works, it is invisible.
For this reason, I am somewhat reluctant to embrace the Ianguage du jour. If I switch to Ruby or Python, I have a different infrastructure. How good is it? For example, how do I render XHTML into PDF? Is it like moving from the United States to Mexico? Fun at first, but then the bad plumbing gets to you? Maybe the JVM will become a part of the infrastructure, and in ten years, I'll call Flying Saucer from JRuby or, more likely, Scala.