Detecting and Addressing Design Smells in Novice Processing Programs

Abstract

Many novice programmers are able to write code that solves a given problem, but they struggled to write code that adheres to basic principles of good application design. Their programs will contain several design smells which indicate a lack of understanding of how to structure code. This applies in particular to degrees in which programming, and by extension software design, is only a small part of the curriculum.This paper defines design smells for Processing, a language for new media and visual arts that is based on Java. This includes language specific smells that arise from the common structure that all Processing programs share. The paper also describes how to detect those smells automatically with static analysis. This tool is meant to support teaching staff with providing feedback to novices on program design.We applied the tool to a large set of student programs, as well as programs from the Processing community, and code examples used by textbooks and instructors. The latter gave a good sense of the quality of resources that students use for reference. We found that a surprising number of resources contains at least some design smell. The paper then describes how to refactor the code to avoid these smells. These guidelines are meant to be practical and fitting the concepts and constructs that are known to first-year students.

Automated Assessment of Learning Objectives in Programming Assignments

“…This goal is not just about using the keyword class, but about structuring the program by grouping related data and actions together. The paper [6] proposes an object-oriented structure for interactive Processing applications and also provides static analysis rules for automated detection of so-called design smells in this structure. If a program is structured into multiple classes and none of the common design smells is detected, the student has likely mastered this objective.…”

“…(b) Relatively few detected design smells in the code patterns. For this metric, Apollo runs the static analysis rules defined in [6] to detect code smells in the program. Because the chance of a small mistake is bigger in a large program, the amount of smells is divided by the number of classes counted for the first metric.…”

“…Finding evidence means to recognise relevant usage of programming constructs and programming patterns. Previous research on static analysis of Processing projects [2,6] successfully adapted PMD 3 . It is also used for Apollo.…”

“…For an initial validation and determining the parameters for converting counted metrics into probabilities, Apollo was used on an old dataset of final student projects used in the evaluation of [6].…”

“…Teaching assistants can leave comments and see each others' comments, which contributes to consistency in giving feedback. Additionally, an automated code checker [6] identifies standard faults, creating suggestions for comments that the teaching assistant can share and discuss with students. This increases the efficiency of giving feedback.…”