Visual Programming Languages: ThingLab

1
Visual Programming Languages ThingLab

• Theuns Verwoerd
• May 1999

2
Ancestry of the ThingLab system

• Sketchpad
• Master/Instance relationships
• Constraints
• Merge points
• Smalltalk
• Object orientation
• Inheritance

3
Origins of ThingLab

• Stanford doctoral thesis by Alan Borning, 1979
• Developed at Xerox PARC in Smalltalk-76
• Aims
• Educational, modelling tool
• Expanding language boundaries
• Extensibility to new domains

4
Core concepts

• Constraints
• Defines relationships between objects
• Multidirectional
• Simplifies component specification
• Implemented as Rule/Satisfaction Method pairs
• Managed automatically by ThingLab
• eg. sum value a value b value

5
Core concepts (2)

• Object orientation
• Multiple inheritance
• Part-whole structure
• Paths to access subsidiary objects
• Definition by example - prototypes
• Merges
• Identifies parts shared between objects

6
User Interface

• Panes
• Class
• Format
• Message
• Argument
• Work space
• Displays prototype image

7
User Interface (2)

• Alternative item views
• Structure
• Prototype values

8
Defining new components

• Done via text source files
• Constraint format

Constraint owner Plus prototype rule 'sum
value (a valueb value)' methods ('sum
primitiveSet.value a valueb value' 'a
primitiveSet.value sum value-b value' 'b
primitiveSet.value sum value-a value')!
9
Some applications
10
More Applications
11
VPL Aspects

• Is it a VPL or a toolkit?
• Abstraction Highly concrete
• Immersive Encourages experimentation, not
  analysis
• Rule (constraint) base
• Usage aimed at novice users, not system
  development

12
VPL Aspects (2)

• Handling complexity
• User-defined composite components
• Problems with large systems
• Representing relationships
• Generally by connection
• Not uniform - sometimes not shown
• Icons vary by component

13
VPL Aspects (3)

• Connection structures
• May form networks
• Hierarchic structure encouraged by composite
  components
• Uses gestures to establish associations
• Animation effects - minimal
• Colour use - non-existent

14
Good points of ThingLab

• Very intuitive for users
• Highly extensible
• Whole/part decomposition simplifies development
• Continuous, immediate feedback

15
Problems with ThingLab

• Installation difficult
• Minimal documentation available
• Fragile - no error recovery
• Scaling problems - time and screen space
• Allows incomprehensible structures
• No concept of time or flow

16
Conclusion

• Attains stated goals well
• Implementation imperfect, but can be fixed
• Serious scaling problems
• Has great potential as a training tool
• Questions