Python is one of the fastest-growing programming languages and currently ranks as the top language in many lists, even recently overtaking JavaScript as the top language on GitHub. Given its importance in data science and machine learning, it is imperative to be able to effectively train LLMs to generate good unit test cases for Python code. This motivates the need for a large dataset to provide training and testing data. To date, while other large datasets exist for languages like Java, none publicly exist for Python. Python poses difficult challenges in generating such a dataset, due to its less rigid naming requirements. In this work, we consider two commonly used Python unit testing frameworks: Pytest and unittest. We analyze a large corpus of over 88K open-source GitHub projects utilizing these testing frameworks. Using a carefully designed set of heuristics, we are able to locate over 22 million test methods. We then analyze the test and non-test code and map individual unit tests to the focal method being tested. This provides an explicit traceability link from the test to the tested method. Our pyMethods2Test dataset contains over 2 million of these focal method mappings, as well as the ability to generate useful context for input to LLMs. The pyMethods2Test dataset is publicly available on Zenodo at: https://doi.org/10.5281/zenodo.14264518

There are possible benefits and drawbacks to chaining methods together, as is often done in fluent APIs. A prior study investigated how Java developers chain methods in over 2.7k open-source projects. That study observed, for the dataset analyzed, that the use of method chaining in Java is popular and seems to be increasing over time. That study however was limited to a smaller sample of Java projects, and it is also not clear if the results generalize to other languages. In this work, we first replicate the prior results by building a similar dataset and our own analysis scripts. We then extend those results by analyzing a much larger dataset of 89k Java projects and generalizing to other programming languages by analyzing 26k Kotlin projects and 98k Python projects. The results show chaining is more popular in Java and Kotlin than Python, chaining use in Kotlin is not growing, and Python sees more use in non-testing code.

Mining Software Repositories (MSR) has recently seen a focus toward ultra-large-scale datasets. Several tools exist to support these efforts, such as the Boa language and infrastructure. While Boa has seen extensive use, in its current form it is not always possible to perform the entire analysis task within the infrastructure, often requiring some post-processing in another language. This limits end-to-end reproducibility and limits sharing/re-use of MSR queries. To address this problem, we use the notion of views from the relational database field and designed a query language and runtime infrastructure extension for Boa that we call materialized views. Materialized views provide output reuse to Boa users, so that the results of prior Boa queries can be easily reused by others. This allows for computing results not previously possible within Boa and provides for increased sharing and reuse of MSR queries. To evaluate views, we performed two partial reproductions of prior MSR studies utilizing Boa's dataset and infrastructure and compare our results to the prior studies. This shows the usability of the new infrastructure, allowing analyses in Boa that were not previously possible as well as providing a previously hand created gold dataset for identifier splitting as a reusable view for other MSR researchers. We also verified the caching behavior using the queries from one of the case studies. The results show that caching works as expected and can drastically improve runtime performance.