Lyra: An Interactive Visualization Design Environment
idl.cs.washington.edu
idl.cs.washington.edu
...yeah, but how exactly? Because it looks about as complicated of a GUI as Tableau...and I don't have enough knowledge of Tableau to compare it against the video, as Tableau's interface is so befuddling that I thank God I stumbled unto web development, as painful as that journey has been, so that I could code my own interactives rather than have to learn Tableau's conventions.
I guess the issue with Lyra is the same with all other programs that claim "custom visualization design without writing any code"...the two desired features, "custom" and "without writing any code"...are, IMHO, at odds with one another. If you want to do anything custom and interactive, you will pretty much have to do something as complicated as code...and pushing a series of buttons and clicking through menus may end up as being as intellectually challenging as just learning programming.
Also, I don't see how the Lyra visualizations are comparable to D3...D3 is amazing because it is a relatively minimalistic framework for coding visualizations...the kind of flexible, expressive visualizations you can do are possible because you are allowed to expressively code them via D3. I don't really see how Lyra (or any GUI) could accomplish that conceptual feat.
One crucial thing to make the GUI simpler than code is direct manipulation. Working with a graphics program, like Inkscape, for example, is much easier for most graphics tasks than writing raw SVGs.
For most tasks in Lyra, you can simply drag-and-drop elements onto the visualization to make it respond the way you want. In advanced cases, true, you do have to navigate menus and buttons. For many of those cases, rather than writing raw D3 code, there's the Vega project. Vega is a declarative format for describing visualizations using a JSON format. Lyra generates Vega specifications for rendering, which can then be modified manually.
Tableau is great for quickly pulling in data and building towards pretty standard charts, then playing around within them to explore the data. It drives you towards relatively simple charts that most people can understand and that's a good thing because most of it's users need significant guidance. And tableau is also a dashboarding platform, but I'm assuming you're talking about just the main visualization builder.
Lyra on the other hand is more of a photoshop like tool for producing unique data graphics. That's where the comparison to custom visualizations in d3 makes sense. d3 isn't about building standard charts (though you can do so) it's a framework for building all sorts of custom rich interactive visualizations. You can do some data exploration in Lyra, but a more constrained tool like tableau, spotfire, etc is easier for that purpose. Where Lyra is interesting IMO is in it's ability to make more custom visualizations without coding - seems like using some kind of grammar of graphics definition under the hood.
Programming languages/frameworks like R, julia, pandas, d3 etc are of course the most flexible tools but they've got a high barrier of entry for a lot of folks who need to produce visualizations and explore data. For the most part this level of vis customization has been restricted to writing a program - lyra feels a lot like a nicer visual ggplot designer in R - so it's nice to see someone making an attempt at bringing some of these capabilities into a gui.
Some feedback if you need it: the UI seems very slow on Chrome, and I don't find it intuitive - there's a lot of dragging and dropping from various places to establish a simple data vis. Ideally you'd like to see what is available, choose it and later play to tweak it, not having to do lots of configuration upfront. Perhaps you could default your tool to do a line chart (or some other kind of early visual feedback ) so the user knows he's on the right track from the beginning?