NullReferenceException with nested parameterized Tests

I’m working with parameterized complex tests (simplified version below). When using a second parameterized test the 'sub’testvalues are passed through from the top level testcase. Which results in a NullPointer regarding my toplevel dataManager.

I already tried to call my testvalue-method directly. This does not work as well. Only instanciating my test object inline seems to work (but not usable because of too much code)

How does building tests influence the object instanciation / execution of before,after steps?

void before_all()
{
    dataManager = new DataManager(system);
    dataManager.Setup();
}

void after_all()
{
    dataManager.Cleanup();
}

void one_test()
{
    new Each<string, RecurrentTestParams3>
    {
        { "Description", new RecurrentTestParams3("Param Description", () => dataManager.GetInputDataArray(), () => ValidateData()) }
    }.Do((configName, data) =>
    {
        context["{0} with {1} on {2}".With(configName, data.Name, system)] = () => doCommonTest(configName, data);
    });
}

private void doCommonTest(string configName, RecurrentTestParams3 datas)
{
    it["Basic Check"] = () =>
    {
        sample.ShouldNotBeNull();
    };

    context["Then Process"] = () =>
    {
        new Each<string, MyDataType>
        {
            //{ "One", datas.InputData()[0]}, --original
            //{ "Two", datas.InputData()[1]}, --original
            //{ "Three", datas.InputData()[2]} --original
            //{ "One", dataManager.getData1()}, --modified
            //{ "Two", dataManager.getData2()}, --modified
            //{ "Three", dataManager.getData3()} --modified
            { "One", new MyDataType() }, -- works
            { "Two", new MyDataType()}, -- works
            { "Three", new MyDataType()} -- works
        }.Do((dataName, data) =>
        {
            it["Work {0}".With(dataName)] = () => operationResult = sample.DoWork(data);
        });

        context["Then Validate"] = () =>
        {
            beforeAll = () => results = dataManager.QueryData();

            data.Validator();
        };
    };
}

[r-weisser]

Yes, I guess I’m just digging into it and this seems to be the best approach. I wasn’t aware of this way of functionality and that I mixed testdata with data needed to build test context.

Regards

[BrainCrumbz]

The way we create parametric tests in my team is:

1. create a kind of blueprint/template of actual test code, with all contexts and befores, its, etc. in place, where the only holes are actual data, say for example string scenario, int limit, double expected.

   context[$"given {scenario}"] = () =>
   {
     before = () =>
     {
       // ...
       someMock.Limit = limit;
       // ...
     };
     
     it[$"returns {expected}"] = () =>
     {
       // ...
       actual.Should().Be(expected);
     };
   }
   

2. Then we define locally some TestDataPoint class that’s just a tuple holding the needed data values:

   class TestDataPoint
   {
     string scenario;
     int limit;
     double expected;
   
     // maybe, a costructor to set values
   }
   

3. move the test code blueprint into some buildContext(TestDataPoint point) method. And iterate over some data set in order to generate tests.

   new []
   {
     new TestDataPoint(/* ... */),
     new TestDataPoint(/* ... */),
     new TestDataPoint(/* ... */),
   
   }.Do(buildContext);
   

(please forgive any syntax error, that was just typed in here)

Less moving parts between test-build and test-run time, just data generated and passed along. HTH