Code Complexity in Introductory Programming Courses

Abstract

Instructors of introductory programming courses would benefit from having a metric for evaluating the sophistication of student code. Since introductory programming courses pack a wide spectrum of topics in a short timeframe, student code changes quickly, raising questions of whether existing software complexity metrics effectively reflect student growth as reflected in their code. We investigate code produced by over 800 students in two different Python-based CS1 courses to determine if frequently used code quality and complexity metrics (e.g., cyclomatic and Halstead complexities) or metrics based on length and syntactic complexity are more effective as a heuristic for gauging students' progress through a course. We conclude that the traditional metrics do not correlate well with time passed in the course. In contrast, metrics based on syntactic complexity and solution size correlate strongly with time in the course, suggesting that they may be more appropriate for evaluating how student code evolves in a course context.

Introduction to the Minitrack on Measurement and Assessment

“…In [2], Petri Inhatola and Andrew Petersen investigate more than 800 student-produced software solutions to introductory python assignments in CS1level programming courses. Their aim was to ascertain if traditional complexity metrics (such as cyclomatic complexity) correlates with student learning outcomes.…”

When Test Cases Are Not Enough: Identification, Assessment, and Rationale of Misconceptions in Correct Code (MC³)

“…However, in CS1, even code that produces the desired outcome can exhibit undesirable characteristics that experienced programmers would typically avoid (De Ruvo et al, 2018;Soloway & Ehrlich, 1984). Examples of such characteristics include higher code complexity resulting from the redundant use of syntactic constructs, such as conditional statements or nested loops (Ihantola & Petersen, 2019;Silva et al, 2021;Ureel II & Wallace, 2019). The sole focus on code correctness may lead students to disregard other important code attributes, such as readability and maintainability, which are crucial for future programmers (De Ruvo et al, 2018;Keuning et al, 2019).…”

Passar nos casos de teste é suficiente? Identificação e análise de problemas de compreensão em códigos corretos

“…Essa utilização promove estudos que visam entender e auxiliar o ensino e a aprendizagem com base em características funcionais [3,6,26,28]. No entanto, em CS1, mesmo códigos corretos podem possuir características indesejáveis que programadores mais experientes não desenvolveriam [13], como, por exemplo, uma maior complexidade resultante da utilização redundante de construtos sintáticos (como demasiados desvios condicionais ou aninhamento de laços) [21,33,40]. Em especial, o foco na corretude pode estimular a não preocupação com fatores de legibilidade e manutenibilidade, essenciais para futuros programadores [13,23].…”