Previous Up Next

5  Tool Overview and Architecture

Input/Processing/Output (black-box view. Figure 2 shows the more general usage of the tool. The input is either (1) a CVS log file —extracted from the CVS version repository with the command cvs log—, or (2) a textual representation of the co-change graph in Rigi Standard Format (RSF) to compute layouts for co-change graphs extracted from other version control systems. To display a previously computed layout, the input can also be (3) a text file containing the layout (LAY).

After the input graph is read, the tool computes the layout for the vertices of the graph. This part is done by an implementation of an energy model and an algorithm that computes a layout that has minimal energy according to the energy model.

At the end, the graph needs to be displayed on the screen, or written to a file. The layout of the artifacts can be produced in three forms. (1) The text file (LAY) can later be read by CCVisu or other tools, such that the tool can be embedded in different environments. (2) The VRML format allows the use of an external VRML viewer (or a web browser with VRML plug-in) to view the layout, and is enabled for 2D as well as 3D layouts. Artifacts are drawn as spheres, and when the mouse pointer moves on the artifacts the name is annotated. (3) The layout can be directly displayed on the screen. This form is the preferred output method for huge graphs, when a VRML viewer is not able to reproduce the layout on the screen. Only two-dimensional layouts are supported. Artifacts are drawn as filled circles, and the names of the artifacts are shown in a separate frame when the mouse pointer moves onto an artifact, or permanently annotated to the artifact via mouse click. (4) Besides the layouts, the tool can also output the extracted co-change graph in RSF format.


Figure 2: CCVisu’s input/output interface

Framework (white-box view). CCVisu is designed as a framework to make improvements and extensions easy, and to enable integration into other reengineering tools. Figure 3 shows the components of the tool. Basically, the input graph is read by a reader component, passed to the layout algorithm, and the output is written by a writer component. The reader interface has currently three implementations: for reading CVS log files, co-change graphs in RSF, and layouts in text format. The writer interface has five implementations so far, for writing co-change graphs in RSF format, layouts in text format, VRML format, SVG format, and for writing the layout directly to the screen.

Using these flexible input/output formats, the tool can be used as a general tool for force-directed graph layout, not only for co-change graphs. To provide more control over the concrete layout computation, the minimizer algorithm and the energy model themselves are also abstract components. Currently, CCVisu includes the Barnes-Hut algorithm [BH86] as minimizer, and the following energy models for the evaluation of layouts: the Fruchterman-Reingold model [FR91], the LinLog model [Noa04a], the edge-repulsion LinLog model [BN05a, Noa04b], and the weighted edge-repulsion LinLog model.


Figure 3: CCVisu’s framework architecture

API. The integration of CCVisu into other tools is possible either by invoking it as a command line tool from a shell, or by invoking the methods of the Java classes using CCVisu’s API.

© Dirk Beyer
Previous Up Next