Library migration is the process of replacing one library with another library that provides similar functionality. Manual library migration is time consuming and error prone, as it requires developers to understand the APIs of both libraries, map them, and perform the necessary code transformations. Large Language Models (LLMs) are shown to be effective at generating and transforming code as well as finding similar code, which are necessary upstream tasks for library migration. Such capabilities suggest that LLMs may be suitable for library migration. Accordingly, this paper investigates the effectiveness of LLMs for migration between Python libraries. We evaluate three LLMs, LLama 3.1, GPT-4o mini, and GPT-4o on PyMigBench, where we migrate 321 real-world library migrations that include 2,989 migration-related code changes. To measure correctness, we (1) compare the LLM’s migrated code with the developers’ migrated code in the benchmark and (2) run the unit tests available in the client repositories. We find that LLama 3.1, GPT-4o mini, and GPT-4o correctly migrate 89%, 89%, and 94% of the migration-related code changes, respectively. We also find that 36%, 52% and 64% of the LLama 3.1, GPT-4o mini, and GPT-4o migrations pass the same tests that passed in the developer’s migration. To ensure the LLMs are not reciting the migrations, we also evaluate them on 10 new repositories where the migration never happened. Overall, our results suggest that LLMs can be effective in migrating code between libraries, but we also identify some open challenges.

Software systems are increasingly relying on deep learning components, due to their remarkable capability of identifying complex data patterns and powering intelligent behaviour. A core enabler of this change in software development is the availability of easy-to-use deep learning libraries. Libraries like PyTorch and TensorFlow empower a large variety of intelligent systems, offering a multitude of algorithms and configuration options, applicable to numerous domains of systems. However, bugs in those popular deep learning libraries also may have dire consequences for the quality of systems they enable; thus, it is important to understand how bugs are identified and fixed in those libraries.Inspired by a study of Jia et al., which investigates the bug identification and fixing process at TensorFlow, we characterize bugs in the PyTorch library, a very popular deep learning framework. We investigate the causes and symptoms of bugs identified during PyTorch’s development, and assess their locality within the project, and extract patterns of bug fixes. Our results highlight that PyTorch bugs are more like traditional software projects bugs, than related to deep learning characteristics. Finally, we also compare our results with the study on TensorFlow, highlighting similarities and differences across the bug identification and fixing process.

Developers heavily rely on Application Programming Interfaces (APIs) from libraries to build their projects. However, libraries might become obsolete, or new libraries with better APIs might become available. In such cases, developers replace the used libraries with alternative libraries, a process known as library migration. Since manually migrating between libraries is tedious and error prone, there has been a lot of effort towards automated library migration. However, most of the current research on automated library migration focuses on Java libraries, and even more so on version migrations of the same library. Despite the increasing popularity of Python, limited research has investigated migration between Python libraries. To provide the necessary data for advancing the development of Python library migration tools, this paper contributes PyMigBench, a benchmark of real Python library migrations.PyMigBench contains 59 analogous library pairs and 75 real migrations with migration-related code changes in 161 Python files across 57 client repositories.

Unit tests play a critical role in improving software quality. However, writing effective unit tests from scratch is difficult and tedious. One way to reduce this difficulty is to recommend existing tests of semantically similar functions. However, modifying the recommended tests manually might still be difficult and tedious. For example, developers have to understand various code elements in the recommended tests to accurately replace them with semantically similar code elements from the target application. One way to mitigate the issue is by developing a test migration or reuse technique that could automatically transform the code elements in the recommended tests and migrate them to the target application. However, to develop such a technique, we first need to identify what types of code transformations are required to successfully migrate the recommended tests. Therefore, in this paper, we first recruit two external participants to create JTestMigBench, a benchmark of 510 manually migrated JUnit tests for 186 methods from five popular libraries. We then analyze the code changes in the migrated tests to create JTestMigTax, a taxonomy of test code transformation patterns. Our contributions provide the necessary foundations to develop automated unit test migration or reuse techniques.