[clang-tidy] Check request: readability-avoid-enable-if

[denzor200]

Needs a check that will find redundant usage of std::enable_if and will suggest more simplified overloading without SFINAE at all.
For example, in 90% usages of std::is_same type trait in a cooperation with std::enable_if we don't really want it.

Look at the sample below...

template<typename T>
std::enable_if<std::is_same<T, int>::value>::type process(T value) {
    // int's implementation..
}

template<typename T>
std::enable_if<!std::is_same<T, int>::value>::type process(const T& value) {
    // generic implementation..
}

...which might be changed to:

void process(int value) {
    // int's implementation..
}

template<typename T>
void process(const T& value) {
    // generic implementation..
}

[llvmbot]

@llvm/issue-subscribers-clang-tidy

Author: Denis Mikhailov (denzor200)

Needs a check that will find redundant usage of std::enable_if and will suggest more simplified overloading without SFINAE at all.
For example, in 90% usages of std::is_same type trait in a cooperation with std::enable_if we don't really want it.

Look at the sample below...

template<typename T>
std::enable_if<std::is_same<T, int>::value>::type process(T value) {
    // int's implementation..
}

template<typename T>
std::enable_if<!std::is_same<T, int>::value>::type process(const T& value) {
    // generic implementation..
}

...which might be changed to:

void process(int value) {
    // int's implementation..
}

template<typename T>
void process(const T& value) {
    // generic implementation..
}

[frederick-vs-ja]

I guess the intent of posted example is to avoid checking (implicit) conversion function of a class to int. Surely, I believe we should always write the latter pattern, but to me the example is possibly workaround for some ill-designed classes.

[PiotrZSL]

"which might be changed to"
No it's not.

void test()
{
    short c = 1;
    process(c);
}

it would cause different function to be called in example above.

[denzor200]

@PiotrZSL you right, the behaviour will be changed indeed. But in the most real cases the bevaviour without enable_if is what a programmer really wants.
Suppose we should think about a compromised solution.