Gradual structuring: Evolving the spreadsheet paradigm for expressiveness and learnability

Spreadsheets are arguably the most used form of programming and are frequently used in higher education to teach fundamental concepts about computation. Their success has shown that they are simple enough for a huge number of end users to learn and use. This is in contrast to traditional programming languages and the high dropout rate from introductory programming and computer science. However in comparison to traditional programming languages and structured modelling, spreadsheets are not expressive, placing a limit on the levels of computational thinking that can be taught using the spreadsheet paradigm. This limitation is imposed by the lack of programming language features and abstractions in the paradigm. Furthermore, more advanced spreadsheet features (e.g. array formulae, lookup formulae, R1C1 syntax) can be difficult to learn and use. This paper discusses the idea of adding language features to spreadsheets, enabling the gradual structuring of free-form spreadsheets to more structured models. We propose that this concept is termed Gradual Structuring, and is analogous to the programming language concept of gradual typing. In this analogy, spreadsheets take the place of dynamic programming and structured modelling of static programming. In programming languages, gradual typing allows dynamic programming to be mixed with static programming. It is our contention that dynamic programming is more learnable while static programming is more expressive and abstract. Gradual typing could be used to mitigate the issues in the teaching of traditional programming. Likewise Gradual Structuring can mitigate the conceptual limits that can be taught using current spreadsheets. The key language feature required to enable Gradual Structuring is the ability to logically group cells together so that a single formula can be applied to the grouped cells. This concept, termed cell grouping diminishes and can even eliminate the need for the ubiquitous and error-prone use of copy-pasted in spreadsheets. Moreover, it makes the structure present in spreadsheet models explicit. Cell grouping requires a cascade of other new languages features. Namely a more expressive referencing style, which in turned requires enabling labels to be moved to the row and column headers, and the hierarchical structuring of these headers. Respectively these language features are termed enhanced referencing and semantic axes. The ongoing research focusses on the usability and learnability of these language features. Spreadsheet applications exist that contain aspects of the features mentioned. However these applications do not enable Gradual Structuring and have taken a mainly technical, not human behavioural, approach to evolving the spreadsheet.