QueryClipse

Abstract: Software developers often search for relevant information in the web in order to solve their encountered programming errors and exceptions. While collecting information using search engines, they face different practical challenges, and one of them is-- choosing an appropriate search query for an error or exception. In this paper, we propose a novel search query recommendation approach for errors and exceptions that analyzes exception details (e.g., exception message, stack traces) and context code of the encountered exception, and recommends a ranked list of search queries. We conduct experiments on the proposed approach using 50 exceptions and associated details (e.g., stack traces, context code), and collect results from three popular search engines (Google, Yahoo and Bing) against the recommended queries. The queries return results with a mean average precision of 55.31% and a recall of 35.23%, and the results solve 80% of the exceptions, which are promising. The experiments followed by a user study also show that the recommended queries by our approach are more effective than traditional queries and queries from three existing approaches in terms of precision, recall, percentage of exceptions solved and pyramid score.