Testing database applications is a complex task since it involves designing test databases with meaningful test data in order to reveal faults and, at the same time, with a small size in order to carry out the testing process in an efficient way. This paper presents an automated approach to generating test data (test relational databases and test inputs for query parameters) for a set of SQL queries, with the aim of covering test requirements as obtained from said queries. The test data generation follows an incremental approach where, in each increment, test data are generated to cover a test requirement by re-using test data previously generated for other test requirements. The test data generation for each test requirement is formulated as a constraint satisfaction problem, where constraints are derived from the test requirement, initial database states and previously generated test data. The generation process is fully automated and supports the execution on complex queries and databases. Evaluation is carried out on a real life application, and the results show that small-size generated test relational databases achieve high coverage scores for the queries under test in a short generating time.

中文翻译：

---

数据库查询的增量测试数据生成

测试数据库应用程序是一项复杂的任务，因为它涉及使用有意义的测试数据设计测试数据库以揭示故障，同时使用小尺寸以有效的方式执行测试过程。本文提出了一种为一组 SQL 查询生成测试数据（测试关系数据库和查询参数的测试输入）的自动化方法，目的是涵盖从所述查询中获得的测试需求。测试数据生成遵循增量方法，其中在每个增量中，通过重用先前为其他测试需求生成的测试数据来生成测试数据以覆盖测试需求。每个测试需求的测试数据生成被公式化为约束满足问题，其中约束来自测试需求，初始数据库状态和先前生成的测试数据。生成过程是完全自动化的，并支持对复杂查询和数据库的执行。在实际应用中进行评估，结果表明，小规模生成的测试关系数据库在较短的生成时间内对被测查询实现了较高的覆盖率。