JarCasting
Overview
Mutant Swarm is a mutation testing framework for Hive SQL built on top of HiveRunner. It enables the identification of areas of SQL code bases that have poor test coverage and consequently may be a source of operational risk.
Usage
You can enable Mutant Swarm on your HiveRunner test suites like so:
- Add
mutant-swarmdependency - Replace the runner or the extension
- Execute test suite
- Locate and view report
Dependency
<dependency>
<groupId>com.hotels</groupId>
<artifactId>mutant-swarm</artifactId>
<version>1.0.0</version>
<scope>test</scope>
</dependency>
JUnit4
HiveRunner JUnit4 tests suites use the com.klarna.hiverunner.StandaloneHiveRunner JUnit4 runner implementation. To enable the Mutant Swarm runner you simply need to replace this with the com.hotels.mutantswarm.MutantSwarmRunner implementation:
@RunWith(MutantSwarmRunner.class)
public class HiveSqlEtlTest {
...
@Test
...
}
JUnit5
To use JUnit5 and the extension model, you will need to have at least version 1.1.0 of Mutant Swarm.
HiveRunner JUnit5 tests classes use the com.klarna.hiverunner.HiveRunnerExtension JUnit5 extension implementation. To enable the Mutant Swarm extension you simply need to replace this with the com.hotels.mutantswarm.MutantSwarmExtension implementation:
@ExtendWith(MutantSwarmExtension.class)
public class HiveSqlEtlTest {
...
@MutantSwarmTest
...
}
You also need to annotate the tests that you want run with Mutant Swarm with @MutantSwarmTest instead of @Test.
Locate and view report
Mutant Swarm currently writes out an HTML report to the target/mutant-swarm-reports folder of your project. The name of the file will be the test class name and its package. So, for example, if we are testing the class MutantSwarmRunnerTest in the package com.hotels.mutantswarm, the report will be called com.hotels.mutantswarm.MutantSwarmRunnerTest_SWARM.html. That way, we will get a different report for each test class.
Motivation
Despite fast-paced innovation in the data processing domain, code developed with SQL-based languages forms a significant part of most organisations' data processing pipelines and ETL applications. The fact that SQL-based engines are often added to even cutting edge frameworks suggests that this trend is set to continue; SQL is easy to learn, is powerful, and has a huge developer base.
However, it has lagged behind the continuous improvements made in the areas of software development best practice and test automation. Unlike code developed in other languages, SQL has limited options for fine-grained automated testing and the analysis of code and test quality. While SQL enables rapid development of critical business systems, it simultaneously leads to code bases that are impervious to testing, introducing potential failure risks that would be deemed unacceptable in any other software development domain.
Mutant Swarm is an important component in a broad approach to redress the balance. It is a tool that identifies deficiencies in test suites that target SQL code, exposing areas of potentially critical business logic that have poor test coverage. By understanding where these deficiencies lie one can begin to evaluate the risk posed, and make informed decisions on where best to focus test development effort.
How it works
- Run tests and report
- Sequence SQL genes
- Generate mutant scripts
- Run tests for each mutant
- Mutation Report
Environment
Mutant Swarm specifically targets code written for the Apache Hive SQL engine, however, the principles it employs could be applied to any SQL engine. The tool works in conjunction with HiveRunner, a unit testing framework for Hive SQL, specifically performing analyses on tests suites developed with it. Generally it is recommended that code bases be modularised. However, coverage insights are arguably even more pertinent for complex and monolithic queries.
Limitations and future work
- Execution time is effectively proportional to the product of the number test cases and the size of the SQL code base under test.
- Project currently includes only a small set of gene matchers and possible mutations, this limits the scope of coverage measurement.
- The responsibility for unit test execution and mutation testing is currently conflated.
- If we want to use JUnit5, we have to change every
@Testannotation in the test class with@MutantSwarmTest. Ideally, we would like to simply place the@MutantSwarmTestannotation at the top of the code and leave the@Testannotations unchanged. We have raised this issue on Stack Overflow.
Contact
Mailing List
If you would like to ask any questions about or discuss MutantSwarm please do so on the HiveRunner mailing list at
https://groups.google.com/forum/#!forum/hive-runner-user
Credits
Conceived and designed by Elliot West, developed by Jay Green-Stevens.
Legal
Copyright 2018-2021 Expedia, Inc.
