Update documentation and examples for generators

docs/generators.md

@@ -1,50 +1,125 @@
 <a id="top"></a>
 # Data Generators
 
-_Generators are currently considered an experimental feature and their
-API can change between versions freely._
-
 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.
-
-How does combining generators and test cases work might be better
-explained by an example:
+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:
 ```cpp
 TEST_CASE("Generators") {
-    auto i = GENERATE( range(1, 11) );
-
-    SECTION( "Some section" ) {
-        auto j = GENERATE( range( 11, 21 ) );
-        REQUIRE(i < j);
+    auto i = GENERATE(1, 2, 3);
+    SECTION("one") {
+        auto j = GENERATE( -3, -2, -1 );
+        REQUIRE(j < i);
     }
 }
 ```
 
-the assertion will be checked 100 times, because there are 10 possible
-values for `i` (1, 2, ..., 10) and for each of them, there are 10 possible
-values for `j` (11, 12, ..., 20).
+The assertion in this test case will be run 9 times, because there
+are 3 possible values for `i` (1, 2, and 3) and there are 3 possible
+values for `j` (-3, -2, and -1).
+
+
+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.
+
+## Provided generators
+
+Catch2's provided generator functionality consists of three parts,
+
+* `GENERATE` macro,  that serves to integrate generator expression with
+a test case,
+* 2 fundamental generators
+  * `ValueGenerator<T>` -- contains only single element
+  * `ValuesGenerator<T>` -- contains multiple elements
+* 4 generic generators that modify other generators
+  * `FilterGenerator<T, Predicate>` -- filters out elements from a generator
+  for which the predicate returns "false"
+  * `TakeGenerator<T>` -- takes first `n` elements from a generator
+  * `RepeatGenerator<T>` -- repeats output from a generator `n` times
+  * `MapGenerator<T, U, Func>` -- returns the result of applying `Func`
+  on elements from a different generator
+
+The generators also have associated helper functions that infer their
+type, making their usage much nicer. These are
+
+* `value(T&&)` for `ValueGenerator<T>`
+* `values(std::initializer_list<T>)` for `ValuesGenerator<T>`
+* `filter(predicate, GeneratorWrapper<T>&&)` for `FilterGenerator<T, Predicate>`
+* `take(count, GeneratorWrapper<T>&&)` for `TakeGenerator<T>`
+* `repeat(repeats, GeneratorWrapper<T>&&)` for `RepeatGenerator<T>`
+* `map(func, GeneratorWrapper<T>&&)` for `MapGenerator<T, T, Func>` (map `T` to `T`)
+* `map<T>(func, GeneratorWrapper<U>&&)` for `MapGenerator<T, U, Func>` (map `U` to `T`)
+
+And can be used as shown in the example below to create a generator
+that returns 100 odd random number:
 
-You can also combine multiple generators by concatenation:
 ```cpp
-static int square(int x) { return x * x; }
-TEST_CASE("Generators 2") {
-    auto i = GENERATE(0, 1, -1, range(-20, -10), range(10, 20));
-    CAPTURE(i);
-    REQUIRE(square(i) >= 0);
+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);
+    }
 }
 ```
 
-This will call `square` with arguments `0`, `1`, `-1`, `-20`, ..., `-11`,
-`10`, ..., `19`.
+_Note that `random` is currently not a part of the first-party generators_.
 
-----------
 
-Because of the experimental nature of the current Generator implementation,
-we won't list all of the first-party generators in Catch2. Instead you
-should look at our current usage tests in
-[projects/SelfTest/UsageTests/Generators.tests.cpp](/projects/SelfTest/UsageTests/Generators.tests.cpp).
-For implementing your own generators, you can look at their implementation in
-[include/internal/catch_generators.hpp](/include/internal/catch_generators.hpp).
+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 `ValueGenerator<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.**
+
+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`:
+
+```cpp
+TEST_CASE("type conversion", "[generators]") {
+    auto str = GENERATE(as<std::string>{}, "a", "bb", "ccc");`
+    REQUIRE(str.size() > 0);
+}
+```
+
+## Generator interface
+
+You can also implement your own generators, by deriving from the
+`IGenerator<T>` interface:
+
+```cpp
+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;
+};
+```
+
+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
+`std::unique_ptr<IGenerator<T>>`.
+
+For full example of implementing your own generator, look into Catch2's
+examples, specifically
+[Generators: Create your own generator](../examples/300-Gen-OwnGenerator.cpp).
+

docs/list-of-examples.md

@@ -14,6 +14,9 @@
 - Report: [TeamCity reporter](../examples/207-Rpt-TeamCityReporter.cpp)
 - Listener: [Listeners](../examples/210-Evt-EventListeners.cpp)
 - Configuration: [Provide your own output streams](../examples/231-Cfg-OutputStreams.cpp)
+- Generators: [Create your own generator](../examples/300-Gen-OwnGenerator.cpp)
+- Generators: [Use variables in generator expressions](../examples/310-Gen-VariablesInGenerators.cpp)
+
 
 ## Planned
 

examples/300-Gen-OwnGenerator.cpp

@@ -0,0 +1,59 @@
+// 300-Gen-OwnGenerator.cpp
+// Shows how to define a custom generator.
+
+// Specifically we will implement a random number generator for integers
+// It will have infinite capacity and settable lower/upper bound
+
+#include <catch2/catch.hpp>
+
+#include <random>
+
+// This class shows how to implement a simple generator for Catch tests
+class RandomIntGenerator : public Catch::Generators::IGenerator<int> {
+    std::minstd_rand m_rand;
+    std::uniform_int_distribution<> m_dist;
+    int current_number;
+public:
+
+    RandomIntGenerator(int low, int high):
+        m_rand(std::random_device{}()),
+        m_dist(low, high)
+    {
+        static_cast<void>(next());
+    }
+
+    int const& get() const override;
+    bool next() override {
+        current_number = m_dist(m_rand);
+        return true;
+    }
+};
+
+// Avoids -Wweak-vtables
+int const& RandomIntGenerator::get() const {
+    return current_number;
+}
+
+// This helper function provides a nicer UX when instantiating the generator
+// Notice that it returns an instance of GeneratorWrapper<int>, which
+// is a value-wrapper around std::unique_ptr<IGenerator<int>>.
+Catch::Generators::GeneratorWrapper<int> random(int low, int high) {
+    return Catch::Generators::GeneratorWrapper<int>(std::unique_ptr<Catch::Generators::IGenerator<int>>(new RandomIntGenerator(low, high)));
+}
+
+// The two sections in this test case are equivalent, but the first one
+// is much more readable/nicer to use
+TEST_CASE("Generating random ints", "[example][generator]") {
+    SECTION("Nice UX") {
+        auto i = GENERATE(take(100, random(-100, 100)));
+        REQUIRE(i >= -100);
+        REQUIRE(i <= 100);
+    }
+    SECTION("Creating the random generator directly") {
+        auto i = GENERATE(take(100, GeneratorWrapper<int>(std::unique_ptr<IGenerator<int>>(new RandomIntGenerator(-100, 100)))));
+        REQUIRE(i >= -100);
+        REQUIRE(i <= 100);
+    }
+}
+
+// Compiling and running this file will result in 400 successful assertions

examples/310-Gen-VariablesInGenerators.cpp

@@ -0,0 +1,72 @@
+// 310-Gen-VariablesInGenerator.cpp
+// Shows how to use variables when creating generators.
+
+// Note that using variables inside generators is dangerous and should
+// be done only if you know what you are doing, because the generators
+// _WILL_ outlive the variables -- thus they should be either captured
+// by value directly, or copied by the generators during construction.
+
+#include <catch2/catch.hpp>
+
+#include <random>
+
+// Lets start by implementing a parametrizable double generator
+class RandomDoubleGenerator : public Catch::Generators::IGenerator<double> {
+    std::minstd_rand m_rand;
+    std::uniform_real_distribution<> m_dist;
+    double current_number;
+public:
+
+    RandomDoubleGenerator(double low, double high):
+        m_rand(std::random_device{}()),
+        m_dist(low, high)
+    {
+        static_cast<void>(next());
+    }
+
+    double const& get() const override;
+    bool next() override {
+        current_number = m_dist(m_rand);
+        return true;
+    }
+};
+
+// Avoids -Wweak-vtables
+double const& RandomDoubleGenerator::get() const {
+    return current_number;
+}
+
+
+// Also provide a nice shortcut for creating the generator
+Catch::Generators::GeneratorWrapper<double> random(double low, double high) {
+    return Catch::Generators::GeneratorWrapper<double>(std::unique_ptr<Catch::Generators::IGenerator<double>>(new RandomDoubleGenerator(low, high)));
+}
+
+
+TEST_CASE("Generate random doubles across different ranges",
+          "[generator][example][advanced]") {
+    // Workaround for old libstdc++
+    using record = std::tuple<double, double>;
+    // Set up 3 ranges to generate numbers from
+    auto r = GENERATE(table<double, double>({
+        record{3, 4},
+        record{-4, -3},
+        record{10, 1000}
+    }));
+
+    // This will not compile (intentionally), because it accesses a variable
+    // auto number = GENERATE(take(50, random(r.first, r.second)));
+    
+    // We have to manually register the generators instead
+    // Notice that we are using value capture in the lambda, to avoid lifetime issues
+    auto number = Catch::Generators::generate( CATCH_INTERNAL_LINEINFO,
+        [=]{
+            using namespace Catch::Generators;
+            return makeGenerators(take(50, random(std::get<0>(r), std::get<1>(r))));
+        }
+    );
+    REQUIRE(std::abs(number) > 0);
+}
+
+// Compiling and running this file will result in 150 successful assertions
+

examples/CMakeLists.txt

@@ -44,6 +44,8 @@ set( SOURCES_IDIOMATIC_TESTS
     110-Fix-ClassFixture.cpp
     120-Bdd-ScenarioGivenWhenThen.cpp
     210-Evt-EventListeners.cpp
+    300-Gen-OwnGenerator.cpp
+    310-Gen-VariablesInGenerators.cpp
 )
 
 # main-s for reporter-specific test sources: