Dataset for the paper "What Fake Data Do Developers Use and Why?" ICSME 2026

Dataset for the paper "What Fake Data Do Developers Use and Why?", submitted to ICSME 2026. This dataset contains both the library and clients data: Faking Library (Python Faker). We aim to study a real-world and relevant faking library. We select the Python faking library Faker due to three main reasons: (1) Python is the most popular programming language nowadays, and Faker is the most popular library to generate fake data in the Python ecosystem; (2) due to its command line usage, Faker can be used by client projects written in any programming language (not only in Python); and (3) Python provides a rich software ecosystem with widely adopted projects to support web development, machine learning, language models, and data analysis, enriching our client-side analysis. Client Projects. To find client projects, we relied on the GitHub search API to collect projects that depend upon the Faker library. Currently, Faker has close to 300K dependent projects. We selected the top-500 most popular Faker dependent projects hosted on GitHub according to the number of stars, a metric primarily adopted in the software mining literature as a proxy of popularity. On the median, the 500 selected projects have 290 stars. The top-3 projects with the most stars are: langchain-ai/langchain (framework powered by large language models, 114K stars), QuivrHQ/quivr (framework for building GenAI, 38K stars), and LAION-AI/Open-Assistant (chat-based assistant, 37K stars). Furthermore, among the selected 500 projects, we find projects belonging to worldwide organizations, including Google, Microsoft, Amazon, Netflix, Airbnb, Redis, Mozilla, Fedora, and Salesforce.