Abstract

Applications of spreadsheets reach from word processing that is
specialized on numeric documents to the solution and visualization of
complex mathematical equations. Despite of an increasing problem
awareness among the users, systematic testing and structured development 
methodologies are not widely applied. Testing is too expensive, and
structured development methodologies do not take into
account that spreadsheet users are end-users, and thus have only
little IT-training. 

Because of the diverse areas of application, many different kinds
of spreadsheets can be identified. Aside from small, but computationally
intensive spreadsheets,
there are also huge spreadsheets with regular structure. Three different
approaches for the visualization of regular structures in spreadsheet
programs are introduced in this thesis. 
Logical Areas, Semantic Classes, and Data Modules,
identify different structural aspects of spreadsheets and effectively 
reduce the complexity induced by the size of huge spreadsheets.
The generated visualizations aim to support the spreadsheet writers to find errors. 

To identify Logical Areas and Semantic Classes the cells' contents,
i.e. the formulas, are examined in order to find regular structures.
The Data Modules approach examines the data flow between cells.
In contrast to other visualization techniques, the identified abstract
units are not limited to adjacent cells with equal formulas, but will also
contain similar formula. It is up to the auditor to state the desired
degree of similarity.
Logical Areas are made up from single cells with similar formulas, but without 
considering their spatial dispersion. Semantic Classes evolved from
Logical Areas. They are made up from regularly recurring regions of
cells, i.e. similar cells with similar neighbors. A Semantic Class
can be considered to consist of sets of cells that fulfill the same kind
of task.
A Data Module is a set of cells that contribute to the same result
of the spreadsheet. Thus, a data module can be considered a set of cells
that cooperate to fulfill a given task.

To frame this work the importance of spreadsheets is pointed out by outlining
the development of spreadsheet systems throughout the last 40 years and
by a short description of different possible applications areas. Due to the
fact, that spreadsheets are still considered to be in the responsibility of
end-users, the vocabulary is still somewhat ambiguous. Hence, this thesis
introduces a well defined terminology for dealing with spreadsheet-related
issues. After the presentation of various spreadsheet error studies,
including one that was carried out as part of this thesis, in order to
show the auditing capabilities of Logical Areas, different
approaches for the definition of the spreadsheet development process are
presented. The theoretical foundations and algorithms for identifications
of Logical Areas, Semantic Classes and Data Modules are
presented, as well as different auditing strategies and a prototype using these
visualization approaches.