I am currently writing unit tests for an embedded application using googles unit test framework. Now my boss got upset that the data I test with (i.e. the values with which I call methods of the class under test) is hard wired in the tests. He requests to have this data read-in from a file. His argument is that it would thus be easier to add another test for a corner case that was previously forgotten. I am not that experienced with unit tests but so far that was not how I did it. So I tried to figure out what would be the best way to do it - even if it is a good idea to do it at all. I quickly came across the DDT (data-driven testing) approach.
The google unit test framework has a feature they call "Value-Parameterized Tests". With that my test fixture becomes a template class and I can pass in parameters. However, I see some problems with this:
I would have imagined that something as mature as the google test framework to make it easier. However, they write
value-parameterized tests come handy [when] you want to test your code over various inputs (a.k.a. data-driven testing). This feature is easy to abuse, so please exercise your good sense when doing it!
Additionally there exists this blogpost TotT: Data Driven Traps, also warning me of (abusing) data-driven unit tests.
So my question comes down to:
I am not really bound to googletest and basically free to choose any framework I'd like, though.
EDIT
I found the following statement in an FAQ entry of the googletest FAQs
Google Test doesn't yet have good support for [...] data-driven tests in general. We hope to be able to make improvements in this area soon.
Google Test is used for unit tests of classes and functions. One writes a test suite for the objects one wants to test, and these and the test suite are compiled and linked to a test program that can be executed and then runs the tests.
gtest-parallel is a script that executes Google Test binaries in parallel, providing good speedup for single-threaded tests (on multi-core machines) and tests that do not run at 100% CPU (on single- or multi-core machines).
Unit tests shouldn't depend on infrastructure 🔗There's no way to test this function without a database connection available at the time of testing. If a new developer clones the project they will need to set up a database before they can successfully run the unit tests.
INSTANTIATE_TEST_CASE_P( LeapYearTests, LeapYearParameterizedTestFixture, ::testing::Values( 1, 711, 1989, 2013 )); Here we call the INSTANTIATE_TEST_CASE_P macro with first with a unique name for the instantiation of the test suite. This name can distinguish between multiple instantiations.
GTest has support for it - but maybe they do not know...
Use testing::ValuesIn - like in this simplified example:
class SomeTests : public TestWithParam<int>
{
public:
};
TEST_P(SomeTests, shouldBePositive)
{
    ASSERT_GT(GetParam(), 0);
}
And - the way how to get values from input stream:
std::ifstream inputValuesFromFile("input.txt");
INSTANTIATE_TEST_CASE_P(FromFileStream,
                        SomeTests,
                        ValuesIn(std::istream_iterator<int>(inputValuesFromFile), 
                                 std::istream_iterator<int>()));
To use more complicated types than "int" you need to write an operator >> for it - like:
struct A
{
    int a;
    std::vector<int> b;
}
std::istream& operator >> (std::istream& is, A& out)
{
     std::size_t bSize;
     if ((is >> A.a) && (is >> bSize))
     {
         out.b.reserve(bSize);
         while (bSize-- > 0)
         {
             int b;
             if (!(is >> b))
                break;
             out.b.push_back(b);   
         }
     }
     return is;
}
Of course - in more complicated cases - consider to use XMl-like (json?) formats and some more specialized iterators than std::istream_iterator<T>.
For XML - like formats - you might consider such scheme (this is very hypothetical code - I do not have any such library in my mind):
SomeLib::File xmlData("input.xml");
class S1BasedTests : public TestWithParam<S1>
{};
TEST_P(S1BasedTests , shouldXxxx)
{
    const S1& s1 = GetParam();
    ...
}
auto s1Entities = file.filterBy<S1>("S1");
INSTANTIATE_TEST_CASE_P(S1,
                        S1BasedTests,
                        ValuesIn(s1Entities.begin(), s1Entities .end()));
// etc for any types S1 you want
If there is no such C++-type-friendly library on the market (I searched for 2 minutes and did not find) - then maybe something like this:
SomeLib::File xmlFile("input.xml");
struct S1BasedTests : public TestWithParam<SomeLib::Node*>
{
   struct S1 // xml=<S1 a="1" b="2"/>
   {
       int a;
       int b;
   };
   S1 readNode()
   {
        S1 s1{};
        s1.a = GetParam()->getNode("a").getValue<int>();
        s1.b = GetParam()->getNode("b").getValue<float>();
        return s1;
   }
};
TEST_P(S1BasedTests , shouldXxxx)
{
    const S1& s1 = readNode();
    ...
}
INSTANTIATE_TEST_CASE_P(S1,
                        S1BasedTests ,
                        ValuesIn(xmlFile.getNode("S1").getChildren()));
                        // xml=<S1s> <S1.../> <S1.../> </S1>
// etc for any node types like S1
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With