An Approach to Automatic Display Layout Using Combinatorial Optimization Algorithms

ing the ADL problem There is a relationship between the problem of chip layout and display layout, since both can be abstracted as a graph-layout problem. It should be noted, however, that this abstraction is simplifying in both cases. A (directed or undirected) graph G = (N,E) consists of a finite set N of nodes and a finite set E of edges (ni,nj). The layout of a graph means the placement of this graph on a given area A according to certain constraints. In the area of chip layout, the nodes correspond to the primitives and the edges to the wires connecting them. The placement area is the chip area itself; constraints might be the minimization of the total wire length and the chip area used and the requirement that wires must not cross. In the area of display layout, the nodes correspond to the objects of the GUI and/or graphical objects that the applications display in their windows; the edges correspond to semantic dependencies between objects (they can but need not be displayed). If the total length of the edges is minimized, the objects that are dependent on each other would be placed together on the screen. We can now give a more formal abstraction of the display layout problem. To abstract a display layout problem, we introduce a graph G = hN,Ed,Eu,Ehj with a set of nodes N and three sets of edges Ed, Eu and Eh. The placement area is given by a rectangular area Ap. Each node ni P N = hn1, %, nnj, in which n = uNu, represents an object that is to be placed on the screen. The nodes have attributes that determine their geometrical dimensions (e.g. width x, length y, area A = xy and width-to-lengthratio x/y) and their position on the placement area Ap. The edges in Ed (Eu) are tuples e = (ni,nj) which are (un)ordered. Consequently, the edges in Ed(Eu) are (un)directed. Each edge ei P Eu < Ed can have certain attributes. They determine the visibility of the edge on the screen, the routeing method to be used (if any), constraints between the two edges connected etc. The edges eh P Eh are used to model the hierarchy of the display layout, i.e. the fact that some objects must be placed within others. They consist of ordered tuples as well. An edge eh = (ni,nj) causes nj to be placed within the area Ai of node ni. It should be noted that there is a root node nr which represents the physical screen, and which therefore has fixed dimensions (its area Ar is equal to Ap, of course). Figure 3 shows the modelling of a display layout as a graph. The attributes of nodes and edges are omitted. On the left-hand side a simple display layout is shown; on the right hand side the corresponding graph G according to the edges in Eh is given. 1187 an approach to automatic display layout Figure 3. Abstraction of a display layout