Linked Views
From InfovisWiki
Contents |
Overview
Multiple view systems provide a means for showing different aspects of either a single data set or comparing similar data sets through two or more views. The power of this approach is greatly enhanced by linking and coordinating the views. Some examples of this include:
- overview and detail view - one window shows context through a map or high level view of the data, the other provides focus through a close-up view.
- brushing and linking - moving the mouse over objects in either view highlights the corresponding object in the other view.
- drill down - one view shows the data graphically, while the other view shows it in textual or numeric form through a table view.
- synchronized scrolling - comparing two versions of source code.
These types of coordination are typically hardcoded into the application, providing functionality without user involvement. However, a much richer and more complex coordination environment is possible, though issues relating to automatic setup and user control are still open research topics. With larger numbers of views, view layout and view management become issues.
Conceptual Models of Coordination
Chris North proposed a simple taxonomy based on two basic classes of actions:
- Select - choose and highlight objects to express interest or target them for other forms of manipulation.
- Navigate - move to focus on objects or display other objects, including loading other objects (e.g. scroll, pan, zoom, slice, rotate, ascend/descend tree, follow link, open file, etc.).
Coordinating a pair of visualizations tightly couples one of these action in the one visualization to another action in the other visualization, resulting in three possible combinations:
- Select <-> select
- Navigate <-> navigate
- Select <-> navigate (navigate <-> select)
For examples of systems that fall into each of these categories, see North's Review of Coordinated-Visualization Systems.
Prior Work
It is difficult to cleanly separate coordination from view management, since some systems (such as spreadsheet-based approaches) base coordination linkage on view organization and some systems automatically generate new linked views for the user. However, the intent here is to focus on the coordination with the view management only providing context. The descriptions below are organized by researcher(s).
Chris North and Ben Shneiderman (Snap-Together Visualization)
Snap-together visualization enables users to dynamically created coordinated visualizations without programming. It is based on the relational database model and uses a drag-and-drop interface to map database relations (tables or query results) into visualizations. Dragging a Snap button between views acts to link them, and selecting available actions for each view from a dialog box specifies the type of link to be established (brushing and linking, overview and detail view, drill down, or synchronized scrolling). Their conceptual model maps visualizations to relations and coordination links to joins. North developed a specification language to describe the coupling between views as paired triples of views, actions, and data items.
Tim Pattison and Matthew Phillips (Model-View-Controller)
Pattison and Phillips provide a generic view coordination architecture, which permits categories of useful view coordination that cannot be implemented using Snap’s conceptual model. Examples of these additional coordination types include linking the appearance of selected elements (beyond the selection itself) in multiple views, or coordinating the layout of these selected elements within the views. Pattison and Phillips base their architecture on the Model-View-Controller (MVC) software pattern. View specification is handled separately from view implementation in the mapping from the data model to the view model. Coordination components, watching for state changes across various levels of the architecture, trigger actions within dependent views whenever changes are detected. Although somewhat different in the details, their architecture can be mapped to the filter-map-render components found in the dataflow model.
Jonathan Roberts (Waltz)
Waltz is an AVS-like visualization system, which provides an intelligent user interface that automates many of the steps in the traditional dataflow model to shorten the setup and provide an iterative environment for repeated cycles of data exploration. The display acts as an information filter, which can output a new dataset as input to the next cycle. The layout and manipulation links between views are automatically generated as the user creates new modules specifying the operations to be performed. Although some of the history of the visual exploration can be traced backward through the modules, a fair amount of screen real estate is devoted to representing the operations instead of the results.
Greg Ross and Matthew Chalmers (HIVE)
HIVE is another data flow system employing a drag-and-drop interface.
References
Nadia Boukhelifa and Peter J. Rodgers. A Model and Software System for Coordinated and Multiple Views in Exploratory Visualization, Information Visualization, 2(4): 258-269, 2003.
Christopher North. A User Interface for Coordinating Visualizations Based on Relational Schemata: Snap-Together Visualization, Ph.D. Dissertation, University of Maryland, College Park, May 2000. (alternate link)
Chris North. Multiple Views and Tight Coupling in Visualization: A Language, Taxonomy, and System. In Proceedings of CSREA CISST 2001 Workshop of Fundamental Issues in Visualization, pages 626-632, June 2001.
Chris North, Nathan Conklin, Kiran Indukuri and Varun Saini. Visualization Schemas and a Web-based Architecture for Custom Multiple-View Visualization of Multiple-Table Databases, Information Visualization, 1(3-4): 211-228, December 2002.
Tim Pattison and Matthew Phillips. View Coordination Architecture for Information Visualization, Proceedings of the 2001 Asia-Pacific Symposium on Information Visualisation - Volume 9, pages 165-169, Sydney, Australia, 2001.
Jonathan C. Roberts. Encouraging Coupled Views for Visualization Exploration, Proceedings of SPIE -- Volume 3643, Visual Data Exploration and Analysis VI, Robert F. Erbacher, Philip C. Chen, Craig M. Wittenbrink, Editors, pp. 14-24, March 1999.
Greg Ross, Alistair Morrison, and Matthew Chalmers. Coordinating Views for Data Visualisation and Algorithmic Profiling, Proceedings of the 2nd International Conference on Coordinated and Multple Views in Exploratory Visualization 2004, pages 3-14, July 2004.
Links
Chris North and Ben Shneiderman. A Taxonomy of Multiple Window Coordinations, University of Maryland Computer Science Dept Technical Report CS-TR-3854, 1997.
