Lets consider following example:
#include <functional>
#include <iostream>
using namespace std;
class Caller {
public:
Caller(function<void()> callback) {
callback();
}
};
main() {
#if defined(ONELINER)
Caller caller = [] { cout << "test"; };
#else
function<void()> fun = [] { cout << "test"; };
Caller caller(fun);
#endif // defined(ONELINER)
}
If we simply try to compile it (with -std=c++11 flag) it will happily finish, and display test
when run. However if we define ONELINER
macro compilation will fail with:
prog.cpp: In function 'int main()':
prog.cpp:17:40: error: conversion from 'main()::<lambda()>' to non-scalar type 'Caller' requested
Caller caller = [] { cout << "test"; };
I understand that this is caused by the fact that there is implicit conversion from lambda to std::function
and then implicit conversion from std::function
to Caller
, and we cannot perform 2 conversions at the same time.
Is it somehow possible to make syntax Class object = lambda;
work? I'm asking because I played recently with writing my own small testing framework for educational reasons and I thought that this:
UNIT_TEST(test_name) {
// test content
};
is much more elegant than
UNIT_TEST_BEGIN(test_name)
// unit test
UNIT_TEST_END()
The former can be achieved with lambdas passed into the UnitTest
constructor. But with problem that I described I had to use dirty workaround like:
#define UNIT_TEST(test_name) \
::std::function<void(::Helper*)> test_name_helper; \
::UnitTest test_name ## _test = \
test_name_helper = \
[&] (::Helper* helper)
and it doesn't look elegant at all. But even if this can be done without lambdas I'm still intrigued whether Class object = lamda;
syntax can be achieved.
Aucun commentaire:
Enregistrer un commentaire