arXiv subjects

Pooya Rostami Mazrae

Publications and source records attributed to Pooya Rostami Mazrae.

5 recordsLinked to original sources

arXiv2026

An Empirical Study of the Evolution of GitHub Actions Workflows

CI/CD practices play a significant role during collaborative software development by automating time-consuming and repetitive tasks such as testing, building, quality checking, dependency and security management. GitHub Actions, the CI/CD tool integrated into GitHub, allows repository maintainers to automate development workflows. We conducted a mixed methods analysis of GitHub Actions workflow changes over time. Through a preliminary qualitative analysis of 439 modified workflow files we identified seven types of conceptual changes to workflows. Next, we performed a quantitative analysis over 49K+ GitHub repositories totaling 267K+ workflow change histories and 3.4M+ workflow file versions from November 2019 to August 2025. This analysis revealed that repositories contain a median of three workflow files, and 7.3% of all workflow files are being changed every week. The changes made to workflows tend to be small, with about three-quarters containing only a single change. The large majority of the observed changes have to do with task configuration and task specification in workflow jobs. We did not find any conclusive evidence of the effect of LLM coding tools or other major technological changes on workflow creation and workflow maintenance frequency. Our findings highlight the need for improved tooling to support fine-grained maintenance tasks, such as a broader adoption of dependency management and AI-based support for ensuring and sustaining workflow security and quality.

cs.SE↗

arXiv2023

The GitHub Development Workflow Automation Ecosystems

Large-scale software development has become a highly collaborative and geographically distributed endeavour, especially in open-source software development ecosystems and their associated developer communities. It has given rise to modern development processes (e.g., pull-based development) that involve a wide range of activities such as issue and bug handling, code reviewing, coding, testing, and deployment. These often very effort-intensive activities are supported by a wide variety of tools such as version control systems, bug and issue trackers, code reviewing systems, code quality analysis tools, test automation, dependency management, and vulnerability detection tools. To reduce the complexity of the collaborative development process, many of the repetitive human activities that are part of the development workflow are being automated by CI/CD tools that help to increase the productivity and quality of software projects. Social coding platforms aim to integrate all this tooling and workflow automation in a single encompassing environment. These social coding platforms gave rise to the emergence of development bots, facilitating the integration with external CI/CD tools and enabling the automation of many other development-related tasks. GitHub, the most popular social coding platform, has introduced GitHub Actions to automate workflows in its hosted software development repositories since November 2019. This chapter explores the ecosystems of development bots and GitHub Actions and their interconnection. It provides an extensive survey of the state-of-the-art in this domain, discusses the opportunities and threats that these ecosystems entail, and reports on the challenges and future perspectives for researchers as well as software practitioners.

cs.SE↗

arXiv2023

An Empirical Study on Data Leakage and Generalizability of Link Prediction Models for Issues and Commits

To enhance documentation and maintenance practices, developers conventionally establish links between related software artifacts manually. Empirical research has revealed that developers frequently overlook this practice, resulting in significant information loss. To address this issue, automatic link recovery techniques have been proposed. However, these approaches primarily focused on improving prediction accuracy on randomly-split datasets, with limited attention given to the impact of data leakage and the generalizability of the predictive models. LinkFormer seeks to address these limitations. Our approach not only preserves and improves the accuracy of existing predictions but also enhances their alignment with real-world settings and their generalizability. First, to better utilize contextual information for prediction, we employ the Transformer architecture and fine-tune multiple pre-trained models on both textual and metadata information of issues and commits. Next, to gauge the effect of time on model performance, we employ two splitting policies during both the training and testing phases; randomly- and temporally-split datasets. Finally, in pursuit of a generic model that can demonstrate high performance across a range of projects, we undertake additional fine-tuning of LinkFormer within two distinct transfer-learning settings. Our findings support that to simulate real-world scenarios effectively, researchers must maintain the temporal flow of data when training models. Furthermore, the results demonstrate that LinkFormer outperforms existing methodologies by a significant margin, achieving a 48% improvement in F1-measure within a project-based setting. Finally, the performance of LinkFormer in the cross-project setting is comparable to its average performance within the project-based scenario.

cs.SE↗

arXiv2022

Bot Detection in GitHub Repositories

Contemporary social coding platforms like GitHub promote collaborative development. Many open-source software repositories hosted in these platforms use machine accounts (bots) to automate and facilitate a wide range of effort-intensive and repetitive activities. Determining if an account corresponds to a bot or a human contributor is important for socio-technical development analytics, for example, to understand how humans collaborate and interact in the presence of bots, to assess the positive and negative impact of using bots, to identify the top project contributors, to identify potential bus factors, and so on. Our project aims to include the trained machine learning (ML) classifier from the BoDeGHa bot detection tool as a plugin to the GrimoireLab software development analytics platform. In this work, we present the procedure to form a pipeline for retrieving contribution and contributor data using Perceval, distinguishing bots from humans using BoDeGHa, and visualising the results using Kibana.

cs.SE↗

arXiv2021

Automated Recovery of Issue-Commit Links Leveraging Both Textual and Non-textual Data

An issue documents discussions around required changes in issue-tracking systems, while a commit contains the change itself in the version control systems. Recovering links between issues and commits can facilitate many software evolution tasks such as bug localization, and software documentation. A previous study on over half a million issues from GitHub reports only about 42.2% of issues are manually linked by developers to their pertinent commits. Automating the linking of commit-issue pairs can contribute to the improvement of the said tasks. By far, current state-of-the-art approaches for automated commit-issue linking suffer from low precision, leading to unreliable results, sometimes to the point that imposes human supervision on the predicted links. The low performance gets even more severe when there is a lack of textual information in either commits or issues. Current approaches are also proven computationally expensive. We propose Hybrid-Linker to overcome such limitations by exploiting two information channels; (1) a non-textual-based component that operates on non-textual, automatically recorded information of the commit-issue pairs to predict a link, and (2) a textual-based one which does the same using textual information of the commit-issue pairs. Then, combining the results from the two classifiers, Hybrid-Linker makes the final prediction. Thus, every time one component falls short in predicting a link, the other component fills the gap and improves the results. We evaluate Hybrid-Linker against competing approaches, namely FRLink and DeepLink on a dataset of 12 projects. Hybrid-Linker achieves 90.1%, 87.8%, and 88.9% based on recall, precision, and F-measure, respectively. It also outperforms FRLink and DeepLink by 31.3%, and 41.3%, regarding the F-measure. Moreover, Hybrid-Linker exhibits extensive improvements in terms of performance as well.

cs.SE↗