Localizing Faults in SQL Predicates

Author(s): Yun Guo, Amihai Motro, Nan Li
Venue: 10th IEEE International Conference on Software Testing, Verification and Validation
Date: 18 May 2017

Data Collection Method: Project Artifact(s)


This paper explores a solution to localizing faults in SQL predicates. According to the paper, fault localization is a technique that finds the locations of faults using tests. The current state of fault localization for SQL queries can only detect that the entire WHERE clause is incorrect. From the authors research on row-based dynamic slicing and delta debugging, the researchers were able to develop a more precise and accurate fault localization tool to improve the state of data-centric applications.

The solution the researchers present is a tool called ALTAR. The tool uses two main techniques dynamic slicing and delta debugging. Dynamic slicing essentially focuses on only the relevant clauses in the SQL query and delta debugging finds the input causes the failures, mutates the data, and determines if the mutated columns caused the fault. After modifying these techniques, the authors were able to successfully detect faults more accurately and precisely than standard testing techniques in the following categories: fault clauses due to one column, multiple faults due to several columns, and general faults.