CS_329m_Project.pdf - Google Drive

increase developer productivity and accessibility by enabling simple trans- lation between natural language and code snippets through a Google-like search application. Rather than take the time to re-implement commonly used code, CodeCrawlr will take a brief description of the desired func- tionality and instantly return multiple viable candidate functions. While there exist services to search code based off of keywords, to the best of our knowledge, there does not currently exist a service for performing code search based on functionality/semantics. CodeCrawlr, contains four main components. First, a data downloading and parsing pipeline, which checks the license of open-source GitHub repositories, downloads and organizes them, and parses the code. Second, an embeddings generation and storage pipeline. Third, a query back-end which produces relevant code given a pro- grammer’s in-progress work or natural-language query. Fourth, a front-end which allows users to submit queries and rece

saved by

• Beining Zhou