This is the final installment of a four-part graphical dissection of the “slabtype” text layout algorithm I developed for Public Secrets. For an introduction to the algorithm, visit The slabtype algorithm, Part 1: Background. To review some calculations that set the stage, visit Part 2: Initial calculations. To get into the real meat of the algorithm, visit Part 3: Iterative line splitting.
In this post, we’ll wrap things up by doing our final layout of the text, followed by the source code for the algorithm. The iterative sequence we explored in the previous installment successfully turned our original text:
into seven separate lines:
We’re almost done. Our next task is to assemble these lines into a slab by scaling them all to an identical pixel width:
And finally, to scale the entire slab to fit inside the original box, allowing for a minimum amount of padding on each side:
And with that, the algorithm has run its course. In conjunction with the treemap algorithm, the slabtype algorithm allows us to dynamically lay out the entire contents of a screen like this
using only a collection of quotes as a starting point. Due to randomization in the input to the treemap algorithm, even identical collections of quotes are never laid out in exactly the same way. Within Public Secrets, you can briefly see the dynamism of the slabtype algorithm in action when a box containing a quote resizes as it moves from one location to another—the text inside shifts around to keep pace with the changing dimensions of its enclosing rectangle.
This algorithm was included in Public Secrets as a method called formatInscription, which is reproduced below (AS 2.0). I hope this has been a useful exercise—I’d love to get your feedback, questions, or suggestions for improvement.
This is the third in a four-part graphical dissection of the “slabtype” text layout algorithm I developed for Public Secrets. For an introduction to the algorithm, visit The slabtype algorithm, Part 1: Background. To review the calculations that set the stage for this post, visit Part 2: Initial calculations.
In this post, we’ll take up the real workhorse of the slabtype algorithm—iterative line splitting—followed by the final layout of the text.
Our initial calculations provided us with a single very important number: 8, our ideal character count per line. This number gives us a target to aim for as we split our text into individual lines. We start by dividing our text into its constituent words:
Next, we execute the following iterative sequence:
Here’s a diagram of how we arrive at our first line using this sequence:
In this example, our two word containers initially hold the words “I and “I WISH. Neither of these exceed the ideal character count per line, so we continue adding words, giving us “I WISH and “I WISH IT. The latter is ten characters long, two characters longer than our ideal character count per line of eight. Since the former is seven characters long, resulting in only a one-character difference between its length and the ideal, we store “I WISH as the contents as the first line of the slab.
Here’s how we arrive at the next line:
Here, we have a very similar case. Neither IT nor IT WAS exceed the ideal character count, so we proceed to IT WAS and IT WAS THAT. IT WAS is only two characters off the ideal, compared to three characters for IT WAS THAT, so we pick the former.
The third line plays out a bit differently. With THAT and THAT SIMPLE, we find that THAT SIMPLE’s character count of eleven is nearer the target of eight than THAT’s character count of four, so we go with the latter.
The fourth line plays out much like the first two:
While the fifth pass chooses the first of the two text containers:
The sixth finds ERASE MY matching the ideal character count per line exactly:
Leaving PAST” as our only remaining word, which becomes a line unto itself.
Since we have exhausted our available words, the iterative part of the algorithm ends, resulting in our original text being split into seven lines:
For the final installment, we’ll turn our attention to laying out the text within the target box (source code will be provided).
This is the second in a four-part graphical dissection of the “slabtype” text layout algorithm I developed for Public Secrets. For an introduction to the algorithm, visit The slabtype algorithm, Part 1: Background.
Execution of the slabtype algorithm breaks down into three phases: initial calculations, iterative line splitting, and final layout. In this post, we’ll tackle the initial calculations that set the stage for the work of the algorithm.
Three constants are used to constrain the results of the initial calculations. The first is the average aspect ratio of a single character of the font we plan to use (Alternate Gothic No. 3, in this case). This is important because it gives us a way to guess how wide a particular line of text is going to be without consuming precious time by actually laying it out. I looked at a selection of characters from the font and estimated the aspect ratio accordingly.
Next, I decided upon a character count for the “ideal line length” of text to be fit into the slab. Since I used the “squarified” variant of the treemap algorithm to generate the enclosing boxes, I could be confident that a fixed value would give good results, but a more truly dynamic implementation would derive the ideal line length from a pixel value for the average height or width of a single character in the ideal point size for the display.
We multiply the average character aspect ratio by the ideal line length to arrive at an aspect ratio for the “ideal line”:
These constants are defined once at the beginning of the program. The calculations which follow are executed each time the algorithm is invoked.
We start with the text which the algorithm is tasked with laying out.
The algorithm is then supplied with the dimensions of the slab (box) into which the text is to be placed:
We divide the width of the box by the aspect ratio of the “ideal line” of text which it is to contain, resulting in a pixel value for the height of this line.
When we divide the height of the box by the height of the ideal line, we arrive at a hypothetical line count—an estimate of how many lines the text is going to have to be broken into to fit snugly within the box.
Since we know the number of characters in our text, we divide it by our hypothetical line count to come up with an “ideal character count” per line of text.
As we will see in the next part, this is a very important value—it becomes a benchmark to help us decide where to split the text into lines.
Animator/designer Alessandro Ceglia pointed out to me a few months back that it’s possible that many users of Public Secrets don’t realize that the piece’s visual presentation is almost entirely algorithmic. Once you get beyond the title screen, all visual composition is handled dynamically, and you’ll never be able to take two identical screenshots of the piece. The designer’s statement for the project describes the two main algorithms used in the project: treemaps (an existing solution dating back to 1991) and an original text-layout algorithm which as of this article is dubbed “slabtype.”
Treemaps (which I fell in love with after seeing Martin Wattenberg’s Map of the Market while I was working at Razorfish) are covered extensively elsewhere (description, history, Wikipedia) and are used to generate the overall layout of boxes for each screen of the piece. In this four-part series of posts I’m going to present a graphical breakdown of the slabtype algorithm responsible for laying out the text and quotations which appear inside the boxes (source code will be provided in the final post).
As the design process for Public Secrets began to lead Sharon Daniel and I back towards her original idea of making the treemap algorithm a central metaphor of the piece, it became clear that we could really use some aesthetically pleasing way of dynamically placing arbitrary amounts of text inside a box. Early designs organized the statements of the incarcerated women as collections of titles, but ultimately these approaches were rejected as too index-like and lacking in emotional power.
A subsequent design hit upon what later became our “split” view, which runs text along an arbitrary rectilinear dividing line between halves of a binary. The dividing line was shown to organize the space on either side into rectangular blocks in which text was laid out along the borders of squares. The idea was intriguing, and I started writing some pseudocode to work out how to fit the text to the border: the program would make a “best guess” at the relative point sizes needed and then refine the guess iteratively. The algorithm was never finalized, however, as it was clear that the type was too hard to read in this configuration.
Further exploration led to a new tack in which the text associated with a particular box was chopped up and resized into a slab that approximated the shape of its enclosing rectangle. This method seemed to add an urgent quality to the quotes and supported an approach in which the quotes themselves were the primary visual experience of the piece—dovetailing nicely with Sharon’s original intent to foreground the voices of the women.
The above screen was assembled by hand in Illustrator and Photoshop, but after Sharon agreed that this approach held promise, I started working on the algorithm itself.
“These machines… They must be generous with our needs. They must be forgiving with our errors. They must be eloquent of ourselves.”
—Grid Farmer Perry, in Chapter Seven of Chroma
Today marks the debut of “Generous Machine,” a freshly minted moniker for this site, and a new domain (generousmachine.com) that points to it. (erikloyer.com will continue to work as well.) I’ve been kicking around the idea of blogging under a title for a few months now, and the right one finally came along over the holidays. For me, this title instinctively evokes my favorite interactive works, and so represents both the best experiences of the past as well as my aspirations for the future. I’ll be returning to developing original content this year—and the “generous machine” will be one standard to be aimed for.
Here’s a list of links to works cited in my recent talk “Storytelling in the Age of Divided Screens” at Gallaudet University.
I’m very happy to announce the launch of “Timeframing: The Art of Comics on Screens,” a new website that explores what comics have to teach us about creative communication in the age of screen media.
To celebrate the launch of Upgrade Soul, here’s a screen shot of an eleven year old prototype I made that sets artwork from Will Eisner’s “The Treasure of Avenue ‘C’” (a story from New York: The Big City) in two dynamically resizable panels.
The last couple of months have seen an uptick in published commentary on Strange Rain, much of it owing to notice the app received at this year’s Modern Language Association conference in Seattle.
Dialogue bubbles huddle together in the Unity authoring environment like backstage theatre performers awaiting their chance to shine in the forthcoming iOS and Android release Upgrade Soul, from Opertoon.