# Data Generators

Introduced in Catch2 2.6.0.

Data generators (also known as data driven/parametrized test cases) let you reuse the same set of assertions across different input values. In Catch2, this means that they respect the ordering and nesting of the TEST_CASE and SECTION macros, and their nested sections are run once per each value in a generator.

This is best explained with an example:

TEST_CASE("Generators") {
    auto i = GENERATE(1, 3, 5);
    REQUIRE(is_odd(i));
}

The “Generators” TEST_CASE will be entered 3 times, and the value of i will be 1, 3, and 5 in turn. GENERATEs can also be used multiple times at the same scope, in which case the result will be a cartesian product of all elements in the generators. This means that in the snippet below, the test case will be run 6 (2*3) times.

TEST_CASE("Generators") {
    auto i = GENERATE(1, 2);
    auto j = GENERATE(3, 4, 5);
}

There are 2 parts to generators in Catch2, the GENERATE macro together with the already provided generators, and the IGenerator<T> interface that allows users to implement their own generators.

Combining GENERATE and SECTION.

GENERATE can be seen as an implicit SECTION, that goes from the place GENERATE is used, to the end of the scope. This can be used for various effects. The simplest usage is shown below, where the SECTION “one” runs 4 (2*2) times, and SECTION “two” is run 6 times (2*3).

TEST_CASE("Generators") {
    auto i = GENERATE(1, 2);
    SECTION("one") {
        auto j = GENERATE(-3, -2);
        REQUIRE(j < i);
    }
    SECTION("two") {
        auto k = GENERATE(4, 5, 6);
        REQUIRE(i != k);
    }
}

The specific order of the SECTIONs will be “one”, “one”, “two”, “two”, “two”, “one”…

The fact that GENERATE introduces a virtual SECTION can also be used to make a generator replay only some SECTIONs, without having to explicitly add a SECTION. As an example, the code below reports 3 assertions, because the “first” section is run once, but the “second” section is run twice.

TEST_CASE("GENERATE between SECTIONs") {
    SECTION("first") { REQUIRE(true); }
    auto _ = GENERATE(1, 2);
    SECTION("second") { REQUIRE(true); }
}

This can lead to surprisingly complex test flows. As an example, the test below will report 14 assertions:

TEST_CASE("Complex mix of sections and generates") {
    auto i = GENERATE(1, 2);
    SECTION("A") {
        SUCCEED("A");
    }
    auto j = GENERATE(3, 4);
    SECTION("B") {
        SUCCEED("B");
    }
    auto k = GENERATE(5, 6);
    SUCCEED();
}

The ability to place GENERATE between two SECTIONs was introduced in Catch2 2.13.0.

Provided generators

Catch2’s provided generator functionality consists of three parts,

ChunkGenerator<T>, RandomIntegerGenerator<Integral>, RandomFloatGenerator<Float> and RangeGenerator<T> were introduced in Catch2 2.7.0.

IteratorGenerator<T> was introduced in Catch2 2.10.0.

The generators also have associated helper functions that infer their type, making their usage much nicer. These are

chunk(), random() and both range() functions were introduced in Catch2 2.7.0.

from_range has been introduced in Catch2 2.10.0

range() for floating point numbers has been introduced in Catch2 2.11.0

And can be used as shown in the example below to create a generator that returns 100 odd random number:

TEST_CASE("Generating random ints", "[example][generator]") {
    SECTION("Deducing functions") {
        auto i = GENERATE(take(100, filter([](int i) { return i % 2 == 1; }, random(-100, 100))));
        REQUIRE(i > -100);
        REQUIRE(i < 100);
        REQUIRE(i % 2 == 1);
    }
}

Apart from registering generators with Catch2, the GENERATE macro has one more purpose, and that is to provide simple way of generating trivial generators, as seen in the first example on this page, where we used it as auto i = GENERATE(1, 2, 3);. This usage converted each of the three literals into a single SingleValueGenerator<int> and then placed them all in a special generator that concatenates other generators. It can also be used with other generators as arguments, such as auto i = GENERATE(0, 2, take(100, random(300, 3000)));. This is useful e.g. if you know that specific inputs are problematic and want to test them separately/first.

For safety reasons, you cannot use variables inside the GENERATE macro. This is done because the generator expression will outlive the outside scope and thus capturing references is dangerous. If you need to use variables inside the generator expression, make sure you thought through the lifetime implications and use GENERATE_COPY or GENERATE_REF.

GENERATE_COPY and GENERATE_REF were introduced in Catch2 2.7.1.

You can also override the inferred type by using as<type> as the first argument to the macro. This can be useful when dealing with string literals, if you want them to come out as std::string:

TEST_CASE("type conversion", "[generators]") {
    auto str = GENERATE(as<std::string>{}, "a", "bb", "ccc");
    REQUIRE(str.size() > 0);
}

Random number generators: details

This section applies from Catch2 3.5.0. Before that, random generators were a thin wrapper around distributions from <random>.

All of the random(a, b) generators in Catch2 currently generate uniformly distributed number in closed interval [a; b]. This is different from std::uniform_real_distribution, which should return numbers in interval [a; b) (but due to rounding can end up returning b anyway), but the difference is intentional, so that random(a, a) makes sense. If there is enough interest from users, we can provide API to pick any of CC, CO, OC, or OO ranges.

Unlike std::uniform_int_distribution, Catch2’s generators also support various single-byte integral types, such as char or bool.

Reproducibility

Given the same seed, the output from the integral generators is fully reproducible across different platforms.

For floating point generators, the situation is much more complex. Generally Catch2 only promises reproducibility (or even just correctness!) on platforms that obey the IEEE-754 standard. Furthermore, reproducibility only applies between binaries that perform floating point math in the same way, e.g. if you compile a binary targetting the x87 FPU and another one targetting SSE2 for floating point math, their results will vary. Similarly, binaries compiled with compiler flags that relax the IEEE-754 adherence, e.g. -ffast-math, might provide different results than those compiled for strict IEEE-754 adherence.

Finally, we provide zero guarantees on the reproducibility of generating long doubles, as the internals of long double varies across different platforms.

Generator interface

You can also implement your own generators, by deriving from the IGenerator<T> interface:

template<typename T>
struct IGenerator : GeneratorUntypedBase {
    // via GeneratorUntypedBase:
    // Attempts to move the generator to the next element.
    // Returns true if successful (and thus has another element that can be read)
    virtual bool next() = 0;

    // Precondition:
    // The generator is either freshly constructed or the last call to next() returned true
    virtual T const& get() const = 0;

    // Returns user-friendly string showing the current generator element
    // Does not have to be overridden, IGenerator provides default implementation
    virtual std::string stringifyImpl() const;
};

However, to be able to use your custom generator inside GENERATE, it will need to be wrapped inside a GeneratorWrapper<T>. GeneratorWrapper<T> is a value wrapper around a Catch::Detail::unique_ptr<IGenerator<T>>.

For full example of implementing your own generator, look into Catch2’s examples, specifically Generators: Create your own generator.

Handling empty generators

The generator interface assumes that a generator always has at least one element. This is not always true, e.g. if the generator depends on an external datafile, the file might be missing.

There are two ways to handle this, depending on whether you want this to be an error or not.

Home