Stress-Minimizing Orthogonal Layout of Data Flow Diagrams with Ports - Graph Drawing - page 319

(a) Layout with layer-based algorithm KLay Layered by Schulze et al.
(b) Layout with the CoDaFlow algorithm presented here
Fig. 1. Two layouts of the same diagram. The result of our method, shown in (b), has
less stress, lower edge length variance, less area, and better aspect ratio.
of drawbacks. First, it employs a strict layering which may result in layouts with
poor aspect ratio and poor compactness, especially when large nodes are present.
Furthermore, the diagrams often have long edges and the underlying structure
and symmetries may not be revealed. A second problem with the approach of
Schulze et al. is that it uses a recursive bottom-up strategy to compute a layout
for nested actors independent of the context in which they appear. This can lead
to bad arrangements with poor aspect ratio and a lack of compactness.
This paper presents a fundamentally different approach to the layout of actor-
oriented data flow diagrams designed to overcome these problems. A comparison
of our new approach with standard layer-based algorithm KLay Layered is shown
in Fig. 1. Our starting point is constrained stress majorization [3]. Minimizing
stress has been shown to improve readability by giving a better understanding
of important graph structure such as cliques, chains and cut nodes [4]. However,
stress-minimization typically results in a quite “organic” look with nodes placed
freely in the plane that is quite different to the very “schematic” arrangement
involving orthogonal edges, a left-to-right “flow” of directed edges, and precise
alignment of node ports that practitioners prefer.
The main technical contribution of this paper is to extend constrained stress
majorization to handle the layout conventions of data flow diagrams. In particu-
lar we: (1) augment the P-stress [7] model to handle ports that are constrained
to node boundaries but are either allowed to float subject to ordering constraints