TypeWriter: Neural Type Prediction with Search-Based Validation
Maintaining large code bases written in dynamically typed languages, such as JavaScript or Python, can be challenging due to the absence of type annotations: simple data compatibility errors proliferate, IDE support is limited, and APIs are hard to comprehend. Recent work attempts to address those issues through either static type inference or probabilistic type prediction. Unfortunately, static type inference for dynamic languages is inherently limited, while probabilistic approaches suffer from imprecision. This paper presents TypeWriter, the first combination of probabilistic type prediction with search-based refinement of predicted types. TypeWriter’s predictor learns to infer the return and argument types for functions from partially annotated code bases by combining the natural language properties of code with programming language-level information. To validate predicted types, TypeWriter invokes a gradual type checker with different combinations of the predicted types, while navigating the space of possible type combinations in a feedback-directed manner. We implement the TypeWriter approach for Python and evaluate it on two code corpora: a multi-million line code base at Facebook and a collection of 1,137 popular open-source projects. We show that TypeWriter’s type predictor achieves an F1 score of 0.64 (0.79) in the top-1 (top-5) predictions for return types, and 0.57 (0.80) for argument types, which clearly outperforms prior type prediction models. By combining predictions with search-based validation, TypeWriter can fully annotate between 14% to 44% of the files in a randomly selected corpus, while ensuring type correctness. A comparison with a static type inference tool shows that TypeWriter adds many more non-trivial types. TypeWriter currently suggests types to developers at Facebook and several thousands of types have already been accepted with minimal changes.
Thu 12 Nov Times are displayed in time zone: (UTC) Coordinated Universal Time change
08:00 - 08:02 Talk | Flexeme: Untangling Commits Using Lexical Flows Research Papers Profir-Petru PârțachiUniversity College London, UK, Santanu Kumar DashUniversity of Surrey, UK, Miltiadis AllamanisMicrosoft Research, UK, Earl T. BarrUniversity College London, UK DOI Pre-print Media Attached File Attached | ||
08:03 - 08:04 Talk | FREPA: An Automated and Formal Approach to Requirement Modeling and Analysis in Aircraft Control Domain Industry Papers Jincao FengEast China Normal University, Weikai MiaoEast China Normal University, China, Hanyue ZhengEast China Normal University, Yihao HuangEast China Normal University, Jianwen LiEast China Normal University, China, Zheng WangBeijing Sunwise Information Technology, China, Ting SuEast China Normal University, China, Bin GuBeijing Institute of Control Engineering, China, Geguang PuShanghai Trusted Industrial Control Platform, China, Mengfei YangChina Academy of Space Technology, China, Jifeng HeShanghai Key Lab of Trustworthy Computing, China DOI | ||
08:05 - 08:06 Talk | Mining Assumptions for Software Components using Machine Learning Research Papers Khouloud GaaloulUniversity of Luxembourg, Luxembourg, Claudio MenghiUniversity of Luxembourg, Luxembourg, Shiva NejatiUniversity of Ottawa, Canada / University of Luxembourg, Luxembourg, Lionel BriandUniversity of Ottawa, Canada / University of Luxembourg, Luxembourg, David WolfeQRA, Canada DOI | ||
08:07 - 08:08 Talk | Mining Input Grammars from Dynamic Control Flow Research Papers DOI | ||
08:09 - 08:10 Talk | TypeWriter: Neural Type Prediction with Search-Based Validation Research Papers Michael PradelUniversity of Stuttgart, Germany, Georgios GousiosFacebook & Delft University of Technology, Jason LiuFacebook, USA, Satish ChandraFacebook, USA DOI Pre-print Media Attached | ||
08:11 - 08:30 Talk | Conversations on Analysis 3 Paper Presentations Khouloud GaaloulUniversity of Luxembourg, Luxembourg, Michael PradelUniversity of Stuttgart, Germany, Profir-Petru PârțachiUniversity College London, UK, Rahul GopinathCISPA, Germany, M: Dan HaoPeking University, China |