how to use enable_if with overloads - c++

enum class enabler{};
template<typename T>
class X {
template<typename std::enable_if<std::is_class<T>::value,enabler>::type = enabler()>
void func();
void func(int a);
void func(std::string b);
};
I have this class with these 3 overloads for func. I need the second/third versions to be available for both class/non-class types, and the first version to be available only for class types. when I tried to use enable_if as above, the class instantiation for non-class types gives compile error.

For SFINAE to work, the template argument must be deduced. In your case, T is already known by the time you attempt to instantiate func, so if the enable_if condition is false, instead of SFINAE, you get a hard error.
To fix the error, just add a template parameter whose default value is T, and use this new parameter in the enable_if check. Now deduction occurs and SFINAE can kick in for non-class types.
template<typename U = T,
typename std::enable_if<std::is_class<U>::value,enabler>::type = enabler()>
void func();
And you don't really need a dedicated enabler type either, this works too
template<typename U = T,
typename std::enable_if<std::is_class<U>::value, int>::type* = nullptr>
void func();

I'm not really sure what you're going for with enabler here, but you can't do what you're trying because the declaration for your member function must be valid since T is not deduced by func. To achieve what you want in adding an extra overload, you can use some moderately contrived inheritance.
struct XBaseImpl {
// whatever you want in both versions
void func(int a) { }
void func(std::string b) { }
};
template <typename, bool> struct XBase;
// is_class is true, contains the extra overload you want
template <typename T>
struct XBase<T, true> : XBaseImpl {
static_assert(std::is_class<T>{}, ""); // just to be safe
using XBaseImpl::func;
void func() { } // class-only
};
// is_class is false
template <typename T>
struct XBase<T, false> : XBaseImpl { };
template<typename T>
class X : public XBase<T, std::is_class<T>{}> { };

You are not enabling or disabling something.
You simply want a compile time error in one specific case.
Because of that you don't require to rely on sfinae, a static_assert is enough.
As a minimal, working example:
#include<string>
template<typename T>
class X {
public:
void func() {
static_assert(std::is_class<T>::value, "!");
// do whatever you want here
}
void func(int a) {}
void func(std::string b) {}
};
int main() {
X<int> x1;
X<std::string> x2;
x2.func(42);
x2.func();
x1.func(42);
// compilation error
// x1.func();
}
Once a SO user said me: this is not sfinae, this is - substitution failure is always an error - and in this case you should use a static_assert instead.
He was right, as shown in the above example a static_assert is easier to write and to understand than sfinae and does its work as well.

Related

C++ Function Template Questions

I've been using C# so long, I have a couple of questions about function templates in C++.
template <typename T>
T max(T x, T y)
{
return (x > y) ? x : y;
}
Why do some examples use typename and other examples use class in the template parameter declaration? What is the difference?
Is there any way to restrict T to a particular type, or to a type that derives from a particular type?
Is there any way for a class to have two methods with the same name, except one is templated and the other is not?
UPDATE:
I appreciate all the answers, but several of them contain examples that I won't compile when I try to apply them to my code.
To clarify question 3, I have the following method:
template<typename T>
std::unique_ptr<T> ExecuteSqlQuery(LPCTSTR pszSqlQuery, UINT nOpenType = AFX_DB_USE_DEFAULT_TYPE);
I would like to declare a variation of this that uses CRecordset as T, so that either of the following statements would be valid:
auto result = db.ExecuteSqlQuery<CCustomerRecordset>(L"SELECT ...");
auto result = db.ExecuteSqlQuery(L"SELECT ...");
Why do some examples use typename and other examples use class in the template parameter declaration? What is the difference?
There is no difference between the two in the template parameter declaration, however they both have additional separate meanings in other contexts. E.g. typename is used to mark dependent names as type names and class is used to introduce a class declaration.
Is there any way to restrict T to a particular type, or a type that derives from a particular type?
Yes, one way is to rely on SFINAE to discard instantiations of types satisfying some condition, often facilitated by std::enable_if, e.g. (using C++14):
template<typename T, typename = std::enable_if_t<std::is_base_of_v<SomeBaseClass, T>>
T max(T x, T y)
{
return (x > y) ? x : y;
}
In the upcoming C++20, there will be support for Concepts, which allow one to write
template<std::DerivedFrom<SomeBaseClass> T>
T max(T x, T y)
{
return (x > y) ? x : y;
}
Is there any way for a class to have two methods with the same name, except one is templated and the other is not?
Yes, this is possible. In overload resolution, if both candidates would be equally well matching, the non-templated one will be preferred.
In this particular context both class and typename mean exaclty the same, there is no difference. class is just a bit shorter :-).
Until C++20 we could try and restrict template arguments using sophisticated template metaprogramming in conjunction with SFINAE technique. Basically, it makes template instantiation fail if the argument does not satisfy some condition. While it's very powerfull approach, it has its drawbacks: increased compile times and very long and unclear error messages.
In C++20 we have a new language feature named concepts, which is aimed to do exactly the same in a simple and straightforward way.
Yes, a function template can be overloaded with a regular function. If the both match, the regular function will be chosen. Note however that in general template overload resolution is quite complicated topic.
Why do some examples use typename and other examples use class in the template parameter declaration? What is the difference?
Historically,
Only typename was allowed for simple template, and class should be used for template template parameter:
template <template <typename> class C> void foo();
with usage such as
foo<std::unique_ptr>();
There are now (C++17) interchangeable in those contexts.
Is there any way to restrict T to a particular type, or to a type that derives from a particular type?
You might do that with SFINAE (which has several syntaxes), and in C++20 with Concepts.
template <typename T>
std::enable_if_t<some_trait<T>::value> foo();
Is there any way for a class to have two methods with the same name, except one is templated and the other is not?
Yes you might have several overloads that way
template <template <class> typename C> void foo();
template <int> void foo();
void foo();
or more simply
template <typename T> void foo(T); // #1
void foo(int); // #2
// Note that foo<int> (#1 with T = int) is different than foo (#2)
Old school C++ used 'class', but we now use 'typename'. You can still use class, but typename is recommended.
Yes, you can restrict types via specialisation..
template<typename T> T foo(T x); //< no implementation in the generic case
template<> T foo<float>(T x) { return x; } //< float is allowed
template<> T foo<double>(T x) { return x; } //< double is allowed
And you can handle derived types as well (and there are a few ways to do this)
#include <string>
#include <iostream>
struct Cow {};
template<typename T>
struct Moo
{
// default to false
template<bool valid = std::is_base_of<Cow, T>::value>
static void moo()
{
std::cout << "No moo for you!" << std::endl;
}
// moo if T is a cow
template<>
static void moo<true>()
{
std::cout << "Mooooo!" << std::endl;
}
};
struct AberdeenAngus : public Cow {};
struct Sheep {};
int main()
{
Moo<AberdeenAngus>::moo();
Moo<Sheep>::moo();
return 0;
}
Yes.
class Foo
{
public:
template<typename T>
T thing(T a) { return a; } //< template
float thing(float a) { return a * 5.0f; } //< function overload
};

avoid pointer-to-member-function for non-class type

I am writing a kind of container class, for which I would like to offer an apply method which evaluates a function on the content of the container.
template<typename T>
struct Foo
{
T val;
/** apply a free function */
template<typename U> Foo<U> apply(U(*fun)(const T&))
{
return Foo<U>(fun(val));
}
/** apply a member function */
template<typename U> Foo<U> apply(U (T::*fun)() const)
{
return Foo<U>((val.*fun)());
}
};
struct Bar{};
template class Foo<Bar>; // this compiles
//template class Foo<int>; // this produces an error
The last line yields error: creating pointer to member function of non-class type ‘const int’. Even though I only instantiated Foo and not used apply at all. So my question is: How can I effectively remove the second overload whenever T is a non-class type?
Note: I also tried having only one overload taking a std::function<U(const T&)>. This kinda works, because both function-pointers and member-function-pointers can be converted to std::function, but this approach effectively disables template deduction for U which makes user-code less readable.
Using std::invoke instead helps, it is much easier to implement and read
template<typename T>
struct Foo
{
T val;
template<typename U> auto apply(U&& fun)
{
return Foo<std::invoke_result_t<U, T>>{std::invoke(std::forward<U>(fun), val)};
}
};
struct Bar{};
template class Foo<Bar>;
template class Foo<int>;
However, this won't compile if the functions are overloaded
int f();
double f(const Bar&);
Foo<Bar>{}.apply(f); // Doesn't compile
The way around that is to use functors instead
Foo<Bar>{}.apply([](auto&& bar) -> decltype(auto) { return f(decltype(bar)(bar)); });
Which also makes it more consistent with member function calls
Foo<Bar>{}.apply([](auto&& bar) -> decltype(auto) { return decltype(bar)(bar).f(); });
In order to remove the second overload you'd need to make it a template and let SFINAE work, e. g. like this:
template<typename T>
struct Foo
{
T val;
//...
/** apply a member function */
template<typename U, typename ObjT>
Foo<U> apply(U (ObjT::*fun)() const)
{
return Foo<U>((val.*fun)());
}
};
Alternatively, you could remove the second overload altogether, and use lambda or std::bind:
#include <functional> // for std::bind
template<typename T>
struct Foo
{
T val;
/** apply a member function */
template<typename U, typename FuncT>
Foo<U> apply(FuncT&& f)
{
return {f(val)};
}
};
struct SomeType
{
int getFive() { return 5; }
};
int main()
{
Foo<SomeType> obj;
obj.apply<int>(std::bind(&SomeType::getFive, std::placeholders::_1));
obj.apply<int>([](SomeType& obj) { return obj.getFive(); });
}
How can I effectively remove the second overload whenever T is a non-class type?
If you can use at least C++11 (and if you tried std::function I suppose you can use it), you can use SFINAE with std::enable_if
template <typename U, typename V>
typename std::enable_if<std::is_class<V>{}
&& std::is_same<V, T>{}, Foo<U>>::type
apply(U (V::*fun)() const)
{ return Foo<U>((val.*fun)()); }
to impose that T is a class.
Observe that you can't check directly T, that is a template parameter of the class, but you have to pass through a V type, a template type of the specific method.
But you can also impose that T and V are the same type (&& std::is_same<V, T>{}).

Checking member function overload existence from a template

Is it possible to check whether a class has a certain member function overload from within a template member function?
The best similar problem I was able to find is this one: Is it possible to write a template to check for a function's existence? As I understand it, this doesn't apply in to the case of checking for overloads of functions.
Here a simplified example of how this would be applied:
struct A;
struct B;
class C
{
public:
template<typename T>
void doSomething(std::string asdf)
{
T data_structure;
/** some code */
if(OVERLOAD_EXISTS(manipulateStruct, T))
{
manipulateStruct(data_structure);
}
/** some more code */
}
private:
void manipulateStruct(B& b) {/** some different code */};
}
My question would be if some standard way exists to make the following usage of the code work:
int main(int argc, const char** argv)
{
C object;
object.doSomething<A>("hello");
object.doSomething<B>("world");
exit(0);
}
The only methods I could think of would be to simply create an emtpy overload of manipulateStruct for struct A. Otherwise the manipulation method could of course also be put into the structs to be manipulated, which would make SFINAE an option. Let's assume both of these to not be a possiblity here.
Is there any way to get code similar to the above one to work? Does something similar to OVERLOAD_EXISTS exist, to let the compiler know when to add the manipulateStruct part to the generated code? Or is there maybe some way clever way to make SFINAE work for this case?
Testing overload existence (C++11)
Since C++11, you can use a mix of std::declval and decltype to test for the existence of a specific overload:
// If overload exists, gets its return type.
// Else compiler error
decltype(std::declval<C&>().manipulateStruct(std::declval<T&>()))
This can be used in a SFINAE construct:
class C {
public:
// implementation skipped
private:
// Declared inside class C to access its private member.
// Enable is just a fake argument to do SFINAE in specializations.
template<typename T, typename Enable=void>
struct can_manipulate;
}
template<typename T, typename Enable>
struct C::can_manipulate : std::false_type {};
// Implemented outside class C, because a complete definition of C is needed for the declval.
template<typename T>
struct C::can_manipulate<T,std::void_t<decltype(std::declval<C&>().manipulateStruct(std::declval<T&>()))>> : std::true_type {};
Here I am ignoring the return type of the overload using std::void_t (C++17, but C++11 alternatives should be possible). If you want to check the return type, you can pass it to std::is_same or std::is_assignable.
doSomething implementation
C++17
This can be done with constexpr if:
template<typename T>
void doSomething(std::string asdf) {
T data_structure;
if constexpr (can_manipulate<T>::value) {
manipulateStruct(data_structure);
}
}
The if constexpr will make the compiler discards the statement-true if the condition evaluates to false. Without the constexpr, the compilation will require the function call inside the if to be valid in all cases.
Live demo (C++17 full code)
C++11
You can emulate the if constexpr behaviour with SFINAE:
class C {
// previous implementation
private:
template<typename T, typename Enable=void>
struct manipulator;
}
template<typename T, typename Enable>
struct C::manipulator {
static void call(C&, T&) {
//no-op
}
};
// can_manipulate can be inlined and removed from the code
template<typename T>
struct C::manipulator<T, typename std::enable_if<C::can_manipulate<T>::value>::type> {
static void call(C& object, T& local) {
object.manipulateStruct(local);
}
};
Function body:
template<typename T>
T doSomething()
{
T data_structure;
// replace if-constexpr:
manipulator<T>::call(*this, data_structure);
}
Live demo (C++11 full code)

Concept checking a function doesn't work with movable-only arguments

I have a function that calls a callback function that accepts a movable-only type (for example unique_ptr).
template <typename Function>
void foo(const Function& function) {
BOOST_CONCEPT_ASSERT((
boost::UnaryFunction<Function, void, std::unique_ptr<Bar>));
auto bar = std::make_unique<Bar>();
...
function(std::move(bar));
}
Trying to compile this code, I get a message that the BOOST_CONCEPT_ASSERT line tries to copy the unique_ptr. If I remove the line, the code works fine. It seems that the Boost.Concept library does not support move semantics. Is there any workaround for this without writing my own concept class (which, incidentally, would not be very simple to support both lvalues and rvalues as their arguments).
That's correct. Unfortunately, UnaryFunction as a concept is written as:
BOOST_concept(UnaryFunction,(Func)(Return)(Arg))
{
BOOST_CONCEPT_USAGE(UnaryFunction) { test(is_void<Return>()); }
private:
void test(boost::mpl::false_)
{
f(arg); // "priming the pump" this way keeps msvc6 happy (ICE)
Return r = f(arg);
ignore_unused_variable_warning(r);
}
void test(boost::mpl::true_)
{
f(arg); // <== would have to have std::move(arg)
// here to work, or at least some kind of
// check against copy-constructibility, etc.
}
#if (BOOST_WORKAROUND(__GNUC__, BOOST_TESTED_AT(4) \
&& BOOST_WORKAROUND(__GNUC__, > 3)))
// Declare a dummy construktor to make gcc happy.
// It seems the compiler can not generate a sensible constructor when this is instantiated with a refence type.
// (warning: non-static reference "const double& boost::UnaryFunction<YourClassHere>::arg"
// in class without a constructor [-Wuninitialized])
UnaryFunction();
#endif
Func f;
Arg arg;
};
Since arg is passed by lvalue, there's no way to get that to work with Boost.Concepts. Directly. You could write a hack though. Since we're just calling checking that f(arg) is valid, we could construct a local type for arg that is convertible to unique_ptr<Bar>. That is:
template <typename Function>
void foo(Function f)
{
struct Foo {
operator std::unique_ptr<int>();
};
BOOST_CONCEPT_ASSERT((
boost::UnaryFunction<Function, void, Foo>));
f(std::make_unique<int>(42));
}
Or more generally:
template <typename T>
struct AsRvalue {
operator T(); // no definition necessary
};
template <typename Function>
void foo(Function f)
{
BOOST_CONCEPT_ASSERT((
boost::UnaryFunction<Function, void, AsRvalue<std::unique_ptr<int>>>));
f(std::make_unique<int>(42));
}
That compiles for me on gcc and clang (though gives a warning on clang about unused typedefs). However, at that point, it may be clearer to just write out your own concept to get it to work. Something like Piotr's would be easiest.
#include <type_traits>
#include <utility>
template <typename...>
struct voider { using type = void; };
template <typename... Ts>
using void_t = typename voider<Ts...>::type;
template <typename, typename = void_t<>>
struct is_callable : std::false_type {};
template <typename F, typename... Args>
struct is_callable<F(Args...), void_t<decltype(std::declval<F>()(std::declval<Args>()...))>> : std::true_type {};
//...
static_assert(is_callable<Function&(std::unique_ptr<Bar>)>{}, "Not callable");
DEMO

Function template specialization with overloads with different number of parameters

Consider the following (invalid) code sample:
// a: base template for function with only one parameter
template<typename T>
void f(T t) { }
// b: base tempalte for function with two parameters
template<typename T1, typename T2>
void f(T1 t1, T2 t2) { }
// c: specialization of a for T = int
template<>
void f<int>(int i) { }
// d: specialization for b with T1 = int - INVALID
template<typename T2>
void f<int, T2>(int i, T2 t2) { }
int main() {
f(true); // should call a
f(true, false); // should call b
f(1); // should call c
f(1, false); // should call d
}
I've read this walk-through on why, in general, partial function template specializations won't work, and I think I understand the basic reasoning: there are cases where function template specializations and overloading would make certain calls ambiguous (there are good examples in the article).
However, is there a reason why this specific example wouldn't work, other than "the standard says it shouldn't"? Does anything change if I can guarantee (e.g. with a static_assert) that the base template is never instantiated? Is there any other way to achieve the same effect?
What I actually want to achieve is to create an extendable factory method
template<typename T>
T create();
which also has a few overloads taking input parameters, e.g.
template<typename T, typename TIn>
T create(TIn in);
template<typename T, typename TIn1, typename TIn2>
T create(TIn1 in1, TIn2 in2);
In order to ensure that all necessary factory methods are present, I use static_assert in the function base templates, so that a compiler error is generated if the create method is called with template arguments for which no specialization has been provided.
I want these to be function templates rather than class templates because there will be quite a lot of them, and they will all use input from the same struct hierarchy, so instantiating 10 factories instead of one comes with some overhead that I'd like to avoid (not considering the fact that the code gets much easier to understand this way, if I can just get it to work...).
Is there a way to get around the problem outlined in the first half of this post, in order to achieve what I've tried to get at with the second half?
In response to iavr:
I could do this with plain overloading, which would (given the templates above) give something like
template<typename TIn2>
A create(bool, TIn2);
template<typename TIn2>
A create(int, TIn2);
if I need two different partial specializations with T = A, TIn1 specified and TIn2 still unspecified. This is a problem, since I have some cases (which are really text-book cases for meta-programming and templates) where I know that, for example, one of the arguments will be a std::string, and the other will be of some type that has a property fields and a property grids, which are of types std::vector<field> and std::vector<grid> respectively. I don't know all the types that will ever be supplied as the second argument - I know for sure that there will be more of them than the ones I currently have implemented - but the implementation of the method will be exactly the same.
While writing up this update, I think I've figured out a way to redesign the implementations so that there is no need for the partial specialization - basically, I do the following to cover the case outlined above:
template<>
A create<A, std::vector<field>, std::vector<grid>>(std::vector<field> fs, std::vector<grid> gs);
and then I have to change the calling signature slightly, but that's OK.
I share your concerns that maybe in this particular case there would be no problem having function template partial specializations, but then again, that's the way it is, so what would be your problem using plain overloading?
// a: base template for function with only one parameter
template<typename T>
void f(T t) { }
// b: base template for function with two parameters
template<typename T1, typename T2>
void f(T1 t1, T2 t2) { }
// c: specialization of a for T = int
void f(int i) { }
// d: specialization for b with T1 = int
template<typename T2>
void f(int i, T2 t2) { }
This also takes less typing and I get this is why you don't want to use function objects (which would have partial specialization).
Here is a simple workaround using a class template specialization:
template <typename, typename...>
struct Creator;
template <typename T, typename TIn>
struct Creator<T, TIn>
{
T call(TIn in)
{
// ...
}
};
template<typename T, typename TIn1, typename TIn2>
struct Creator<T, TIn1, TIn2>
{
T call(TIn1 in1, TIn2 in2)
{
// ...
}
};
template <typename R, typename... Arguments>
R Create(Arguments&&... arguments)
{
return Creator<R, Arguments...>::call(std::forward<Arguments>(arguments)...);
}
If you don't want overloading, and want to be able to specialize from a separate file, then I think you should base it on the solution on the link from your question. It involves making a static method on a class that you specialize. From my reading of the question, you're only interested in specializing on the T, not on the number of arguments, which you intend to forward. In C++11, you can do the following:
#include <iostream>
#include <utility>
using namespace std;
template<typename T>
struct factory_impl;
// Left unspecified for now (which causes compliation failure if
// not later specialized
template<typename T, typename... Args>
T create(Args&&... args)
{
return factory_impl<T>::create(std::forward<Args>(args)...);
}
// Note, this can be specified in a header in another translation
// unit. The only requirement is that the specialization
// be defined prior to calling create with the correct value
// of T
template<>
struct factory_impl<int>
{
// int can be constructed with 0 arguments or 1 argument
static int create(int src = 0)
{
return src;
}
};
int main(int argc, char** argv)
{
int i = create<int>();
int j = create<int>(5);
// double d = create<double>(); // Fails to compile
std::cout << i << " " << j << std::endl;
return 0;
}
Live example http://ideone.com/7a3uRZ
Edit: In response to your question, you could also make create a member function of a class, and pass along some of that data with the call or take action before or after
struct MyFactory
{
template<typename T, typename... Args>
T create(Args&&... args)
{
T ret = factory_impl<T>::create(data, std::forward<Args>(args)...);
// do something with ret
return ret;
}
Foo data; // Example
};