Detecting Problems in the Database Access Code of Large Scale Systems - An Industrial Experience Report

Database management systems (DBMSs) are one of the most important components in modern large-scale systems. Thus, it is important for developers to write code that can access DBMS correctly and efficiently. Since the behaviour of database access code can sometimes be a blackbox for developers, writing good test cases to capture problems in database access code can be very difficult. In addition to testing, static bug detection tools are often used to detect problems in the code. However, existing bug detection tools usually fail to detect functional and performance problems in the database access code. In this paper, we document our industrial experience over the past few years on finding bug patterns of database access code, implementing a bug detection tool, and integrating the tool into daily practice. We discuss the challenges that we encountered and the day-to-day lessons that we learned during integrating our tool into the development processes. Since most systems nowadays are leveraging frameworks, we also provide a detailed discussion of five framework-specific database access bug patterns that we found. We hope to encourage further research efforts on framework-specific detectors, instead of the current research focus on general programming language bug patterns and associated detectors.