The practice of software detailed design of graduating students : A family of experiments.

Abstract

Software design is one of the essential components to ensure the success of a software system. It has two main activities: architectural design and detailed design. During architectural design, high-level components are structured and identi ed. During detailed design, every component is speci ed in detail. Several authors consider design is a complex discipline for undergraduate students to understand, and success (i.e., building a good design) seems to require a certain level of cognitive development that few students achieve. Normally, students do not manage to produce a good software design. Some of the problems detected are lack of consistency between design artifacts and code, incomplete designs, and the lack of understanding of what kind of information to include when designing software. The general objective of this thesis is to contribute to the knowledge of how graduating students practice software detailed design. Speci cally, we conducted a family of experiments in the context of a course at the School of Engineering of Universidad de la Rep ublica, in Uruguay. The family of experiments is composed of 3 sub-families of experiments: Baseline, Template, and Habitual. Baseline experiments are made up of an initial experiment (executed in 2012) and two replicates (executed in 2013 and 2014). Template experiments are made up of an initial experiment (executed in 2015) and two replicates (executed in 2016 and 2017). Finally, Habitual experiments consist of an initial experiment (executed in 2018) with one replicate (executed in 2021). Through the family of experiments, we studied the detailed design practice of graduating students from di erent points of view. The results indicate that our graduating students do not take time to think of a solution (design) before coding. They spend at least three times less time designing than coding. In addition, we introduced design templates with the intention that they would be a tool to help them in the design task. However, although the design time signi cantly increases, the quality of the software, measured in the unit testing phase, does not change. Lastly, the analysis of the design representation delivered by the students reveals simple, basic designs with little elaboration.