Post on 30-Jun-2015
DEPENDENCE TRACING TECHNIQUES FOR SPREADSHEETS: AN INVESTIGATION
SOHON ROYFELIENNE HERMANS
The earth as we see in plain sight
How it really is if we consider flight paths
Interconnected!
Spreadsheet in plain sight
How the cells are interconnected
Why bother about dependence?
“the most important information needs of professional spreadsheet users concern the structure of the formula dependencies” – Hermans et al. in 2011
20% problemsInformal Survey
Necessity is the mother of inventions
But getting your paper accepted and published is the father?!
The gap
Not really the best way
Definitely not the best way
The gap
Hypothesis
Better dependence tracing techniques or tools
developed under scientific research is not adequately
making it across to the industry to benefit real-life
spreadsheet users.
Why the bottleneck between research and industry?
A metrics framework to measure user friendliness and effectiveness?
Any opportunity of improvement in the dependence tracing context?
Review of existing research papers
•Relevance to the topic•Citations and references in
related papers•Distinct and unique
approach• Theory demonstrated
through tool/prototype
Limited scope for this paper
SELECT
The papers
Fluid VisualizationIgarashi et al. 1998
3D Interactive VisualizationShiozawa et al. 1999
Visual Checking with VBAChen et al. 2000
Spreadsheet Visualisation with JavaBallinger et al. 2003
Leveled Dataflow DiagramsHermans et al. 2011
Findings
5 of 5
Innovative Ideas & Tool/Prototypes
Findings
2-Defunct 1-Subject to Compatibility Issue2-Current (Java, C#)
Technology
Defunct Subject to Compatibility Current
Performance Testing
Tested Unrealistic Test Not Tested
Findings
2-Tested 1-Unrealistic Test (400 cells)2-Not Tested
User Testing
User Study No User Study
Findings
1 of 5
Exposure to the Industry
Available to the Industry Never exposed to the Industry
Findings
1 of 5
Why the bottleneck between research and industry?
Lack of motivation?
Why the bottleneck between research and industry?
Difficult to scale up performance?
Difficult to test?
Performance Testing
Tested Unrealistic Test Not Tested
Findings
2-Tested 1-Unrealistic Test (400 cells)2-Not Tested
Why the bottleneck between research and industry?
Difficult to scale up performance?
Difficult to test?
A metrics framework to measure user friendliness and effectiveness?
Apparently none.
User Testing
User Study No User Study
Findings
1 of 5
A metrics framework to measure user friendliness and effectiveness?
To define and measure useability of spreadhsheet
tools.
Any opportunity of improvement in the dependence tracing context?
@sohonroyS.Roy-1@tudelft.nl
http://spreadsheetlab.org/http://sohonroy.wordpress.com/
Paper Published In
Technique PrototypePrototype Comments
Industrial Relevance
1.1998 IgarashiFluid Visualization & Animation
Pad++ (now defunct) and Python for UNIX
Performance drastically worsens after 400 cells.No user study.
No implementation plan given. Some ideological influence can be observed in MS Excel 2007 onwards.
2.1999 Shiozawa3D Interactive lift-up operation
OpenGL APIs and SLSC for UNIX
None available. None.
3. 2000 Chen
Visual indications superimposed on Excel sheets like arrows, highlights, shading.
VBA plug-in for Excel.
None available.
No standardized definite influence. However similar techniques are used in ad hoc basis by expert spreadsheet users on individually or internally to organizations.
4.2003 Ballinger
Different types of graphs and trees depicting spreadsheet dependency structure.
Java tool for Excel workbooks.
Successfully tested on a corpus of 259 Excel workbooks of undisclosed source.No user acceptance or comprehension study.
Authors apparently did not carry forward their work on this project.
5.2011 HermansLeveled dataflow diagrams
C# toolkit that accepts Excel workbooks for input.
Extensive user testing with 27 industrial spreadsheet users in a financial asset management company.
Is currently a part of service offered by the company Infotron.nl