The InfoVis Toolkit

1
The InfoVis Toolkit

• Jean-Daniel FeketeINRIA Futurs/LRI,
  FranceJean-Daniel.Fekete_at_inria.frhttp//www.lri.
  fr/fekete/InfovisToolkit

2
Goals

• Simplify the integration of Information
  Visualization techniques in Java/Swing
  applications
• Simplify the creation and experimentation of new
  visualization techniques
• Simplify the creation and experimentation of new
  interaction techniques
• Provide a support for teaching InfoVis
• Provide a base for InfoVis student projects
• Scales well in term of space and speed

3
Internal Structure

• Data structures are composed of Columns
• Columns contain homogeneous data
• Contain metadata
• Support notification
• Support formatting for input and output
• Support missing values
• Efficient in speed and memory usage
• Data structures are based on tables
• Attributes and topology explicitly represented as
  columns
• Trees
• Columns containing the parent, 1st child, next
  child, last child
• Graphs (two tables for vertices and edges)
• In vertex table Outgoing edge list, ingoing edge
  list
• In edge table next edge on outgoing list, next
  edge in ingoing, in and out vertices

4
Results

• Article at InfoVis 04 about the Toolkit and a
  user evaluation
• Technical report about performance issues
• Several projects
• Monitoring of constraint-based systems
• Software testing
• Visualization of familiar data (new Desktop)

5
Problems

• Separation of concerns
• Data structure for graphs is complex when
  scalability and extensibility are required
• Tulip has a good model
• When a good framework exists, implementing parts
  is long, tedious and hard
• Visualizations (layouts)
• Rendering (labels, arrow ends, arcs shapes)
• Components for magic-lenses (Fisheyes,
  excentric labels)
• Components for Dynamic Queries (sliders et al.)
• Components to configure and control
  visualizations
• Readers and writers
• Documentation

6
Is there several ways to do it?

• Looking at implementations, 2 ways
• The good old object oriented way
• The column-oriented way
• The goals are the same for most projects
• Same scales
• Same platforms
• Same languages (almost)
• Should I continue by myself?