Automatic Identification of Code Changes That Require Extensive Testing

点击翻译按钮获取中文摘要

NATURE AND IMPORTANCE OF THE WORK TO BE DONE: Software Quality Assurance (SQA)**activities are performed to ensure that changes to software systems do not introduce regression in software**system properties (e.g., functionality, performance). SQA is a resource-constrained activity-exhaustive**verification of all of the changes that a software organization produces is not practically feasible. SQA**resources should be redirected from low-risk changes to high-risk ones in order to minimize the likelihood of**regressions in software system properties (e.g., bugs) from impacting user experience. At Shopify, additional**SQA resources are allocated to changes that the development team deems to be risky. Yet these decisions may**be error-prone, since "gut feel" is often involved, resulting in a suboptimal allocation of SQA resources. In this**project, we will leverage Shopify's historical development data to support SQA resource allocation decisions.**ANTICIPATED OUTCOMES: In a nutshell, we will train predictive models that will estimate a risk score for**each code change. These predictive models will be integrated into the Shopify code change process.**RELEVANCE TO THE COMPANY: This historical perspective has the potential to transform the way that**Shopify performs SQA. These integrated risk scores will help developers to make data-informed decisions**about the changes to which additional SQA resources should be allocated.**HOW CANADA WILL BENEFIT: This project will provide an opportunity for two master's students (Mrs.**Farida El Zanaty and Mr. Christophe Rezk) to be trained in large-scale statistical regression and applied**machine learning to solve real-world industrial problems. These are invaluable skills that will help to fill the**growing demand for data scientists in the Canadian public and private sectors.