Fire Drill Anti-Pattern Dataset
收藏资源简介：
Fire Drill Anti-Pattern Dataset 是 9 个真实世界软件项目的集合，用于检测具有基本事实、问题跟踪数据、源代码密度、模型和代码的 Fire Drill 反模式。这些数据应该有助于检测 Fire Drill 反模式的存在。它包括数据、基本事实、代码和笔记本。该数据支持两种不同的 AP 检测方法：a) 通过问题跟踪数据，b) 通过底层源代码。因此，该软件包包括以下内容： 问题跟踪数据中的防火练习：每个项目是否以及如何展示防火练习 AP 的基本事实，范围为 [0,10]。这是由两个单独的评估者决定的，他们也达成了共识。每个项目的第一种方法的指标系数。每个项目的详细问题跟踪数据：发生了什么以及何时发生。每个项目的时间日志。源代码数据中的防火练习：三份技术报告，记录了如何将描述转换为可检测模式，并使用该模式检测存在并对其进行评分（类似于评级）的开发方法。还包括有关如何将活动分配给各个提交的报告。九个项目中每个提交的源代码密度数据（指标）作为单独的数据集。代码：存储所有代码、模型、笔记本和预计算结果的存储库快照，以实现最大的可重复性（代码是用 R 编写的）。
The Fire Drill Anti-Pattern Dataset is a collection of 9 real-world software projects designed for detecting Fire Drill anti-patterns, with accompanying ground truth, issue tracking data, source code density, models and code. This dataset facilitates the detection of Fire Drill anti-patterns. It includes data, ground truth, code and notebooks. The data supports two distinct Fire Drill anti-pattern detection approaches: a) via issue tracking data, and b) via underlying source code. Accordingly, this software package contains the following components: 1. Fire Drill practices in issue tracking data: - Ground truth on whether and how each project exhibits Fire Drill anti-patterns, with a score range of [0, 10]. This was determined by two independent evaluators who reached a consensus. - Metric coefficients for the first detection approach of each project. - Detailed issue tracking data for each project, documenting what occurred and when it occurred. - Time logs for each project. 2. Fire Drill practices in source code data: - Three technical reports detailing the development methodology for translating anti-pattern descriptions into detectable patterns, using these patterns to detect their existence and assign scores (similar to ratings). Reports on how to allocate activities to individual commits are also included. - Source code density data (metrics) for each commit across the nine projects, provided as separate datasets. 3. Code: A repository snapshot storing all code, models, notebooks and precomputed results to maximize reproducibility. The code is written in R.
