This article focuses on generating the junit compatible test report that can be consumed by a CI Server.

We collect the result and log of our tests with a dedicated process and then generate the XML based test report. The related junit5 XSD Schema can be found here.

The XML document starts with a testsuite tag containing some summary info:

<testsuite name="TestSuitName" tests="NumberOfTests" failures="ErrorCount" skipped="0" errors="0">

</testsuite>

Inside the testsuite tag we can have several test cases which are constructed like so:

In case the test is a success:

  <testcase name="TestName" classname="ClassName" package="PackageName" time="ExecutionTimeSeconds">
    <system-out>
      <![CDATA[ 
        Execution Log Text 
      ]]>
    </system-out>
  </testcase>

In case the test result is an error:

  <testcase name="TestName" classname="ClassName" package="PackageName" time="ExecutionTimeSeconds">
    <failure type="value" message="ERROR Found"/> <!-- ONLY THIS BIT IS DIFFERENT -->
    <system-out>
      <![CDATA[ 
        Execution Log Text 
      ]]>
    </system-out>
  </testcase>

What follows is a sample implementation with Kettle/PDI/Project Hob. I’d guess that there are many implementations of this already around. The one shown below is mainly based on what Beppe Raymaeker provided for Hitachi Vantara’s official recommendation:

• Framework Repo
• Example Code Repo
• Example Config Repo

You can download the example yourself and go through it, so I will discuss it only on a high level here:

This PDI process (transformation) will load all the tests defined in the test-suite.xml file and then execute each one of them: Each test can define a PDI job or PDI transformation as the test subject. In the below screenshot we can see the sub-transformation which actually executes the job-to-be-tested:

All the execution stats get collected and propagated to the parent job. Once all tests are executed, the parent job constructs the JUnit based test report for all the tests.