The Recurring Rainfall Problem

Abstract

Soloway's Rainfall problem, a classic benchmark in computing education research, has proven difficult for many CS1 students. Rainfall tests students' abilities at plan composition, the task of integrating code fragments that implement subparts of a problem into a single program. Nearly all prior studies of Rainfall have involved students who were learning imperative programming with arrays. In our recent multi-university study, students learning functional programming produced atypical profiles of compositions and errors on Rainfall. The differences have interesting implications for the questions and methodologies we use to study plan composition.

The talk will also discuss pedagogic uses of peer review. While some studies have explored peer review of code, none have considered review of test suites. We present an overview of observations and results from a multi-course study in which students peer reviewed test suites prior to submitting implementations. Our observations prompt more general thoughts on using peer review to target broad objectives for computing education.