Bear in mind that exporting Processing output would leave you with a flat file, instead of manipulable objects. As you say, this sort of heavy-entropy-loss operation might well lead to “argh, I have to start again” situations downstream from that point. I think a scripting engine in a graphics package is definitely the way to go.

I really feel like this calls for an interface layer to be written. You curate your data source, and then you open your addon/bunch of scripts (possibly in Illustrator or Inkscape, they both have great scripting environments by the looks of things but I’m not sure how far they can go in terms of creating GUIs and complex realtime logic) and point it at that source. You are then able to attach data in meaningful ways to, shall we say, “draw tools”, which you can then attach behaviours to (everything from what to draw to how to label it, colours mapped to integer values, etc). Something like aesthetically pleasing, non-interfering labels that intelligently place themselves is actually not too tricky to achieve programmatically, and once written could be added to a behaviour toolbox. You’d eventually build up enough tools to bring whatever you imagine to life.

I want to emphasize that I feel that the human element is very important, and that layout design should always be down to a… designer – just that extensive programmatic support would take so much of the tedium out of creating insightful infographics.

That sort of tool is out of scope for both of us, I’m guessing, ‘though it’d make a great project for someone at the MIT media lab with time to kill :).

I also have some ideas about just straight-up using processing, or at least javascript, to make “brochures”, if you will; essentially pictures like these that use live data feeds. There are some examples of this around (in fact, it’s not an original idea at all; remember the flash digg-data-wheel-thing?) and I think there is very exciting potential here for all sorts of data analysis.

I’m going to do some intense investigation and dabbling in graphics-package-script and web tool draw functions over the next few weeks. I’ll let you know how I progress, if you’re interested.

If I was going to do this all again (and I probably will, but for different kinds of data) I would look into Illustrator’s scripting system. I don’t really know much about it, but it might help with this kind of thing.

The other option is to generate the components of the visualization using Processing, export to a PDF, and then pull that into Illustrator. PDFs are functionally Illustrator files, so you retain full editing ability on them. So I could have generated all the circles and text directly from my spreadsheet and then done the layout by hand. In the end, I think all the real layout/design work has to be done by hand. Trying to write something that automated the placement of the callout text and placed the bubbles relative to each other would be a nightmare. So I kinda like this strategy, where you just use another tool to generate the pieces of the viz and then assemble them by hand. The problem with that is updating stuff – if you do 90% of the layout then realize you need to make a global change, you’re screwed and have to do much of the layout over again.

Do you have any suggestions? I feel like this project fell into a weird middle ground where it wasn’t computational enough to generate it all with Processing, but was still annoying to do by hand.

Damn, now I want to play with inkscape.