Making Structured Graphics and Constraints Practical for Large-Scale Applications

Abstract

Providing a structured graphics model and a constraint system makes the programming of graphical applications significantly easier. In a structured graphics model, each graphic element on the screen is represented by a real object in the object system, while a constraint system automatically maintains relationships among the objects. Although many research systems and a few commercial environments provide structured graphics and constraints, none scale up to large interfaces with thousands of objects and thousands of constraints. This paper presents four techniques to overcome this problem: automatic elimination of constraints that depend only on values that do not change, layout hints to help with refresh and hit detection, "virtual aggregates" that only pretend to allocate objects for their components, and compiling composite aggregates into a single object with a complex draw method. These have been implemented as part of the Garnet environment, and we have demonstrated that together they allow structured graphics and constraints to scale up to applications with tens of thousands of objects and constraints without compromising the ease-of-use.

Optimizing toolkit-generated graphical interfaces

“…Constant propagation has been reported on elsewhere [12,15], and thus will not be addressed further in this paper.…”

“…Traditional methods for reducing overhead involve (1) compiling constraints into plans [6,17] and (2) using constant propagation [12,15].…”

“…Glyphs are extremely lightweight objects that only store information which cannot be computed by other means [5]. Virtual aggregates attempt to eliminate objects completely by constructing objects only when they are demanded by an operation, and then reclaiming the objects once the operation finishes [15]. However, these approaches only partially solve the problem: They reduce the storage cost, but do not eliminate the expensive intersection tests that must be performed at runtime to redraw the individual objects.…”

“…However, this strategy only partially remedies the problem because it is still necessary to 1) dynamically detect damaged areas of the display, 2) find and redisplay the objects that intersect the damaged areas, and 3) update the display list as objects are modified, created, or deleted. Finally, some toolkits, such as Garnet [11,15], allow the programmer to explicitly specify semantic information about an application.…”

Using model dataflow graphs to reduce the storage requirements of constraints

“…7.2.1 Constant Propagation. Constant propagation allows constraints whose parameters are all constants to be evaluated once and replaced with the constant result [Maloney et al 1989;Myers et al 1994]. Constant propagation works well in a system where many slots have fixed values.…”