I'm pretty new in move and lvalue semantics. And I have the impression I'm doing it wrong. Here
the code I want to be able to write once FunctContainer
is implemented:
std::function<double(double)> f = [](double x){return (x * x - 1); };
FunctContainer fc1 = FunctContainer(f);
FunctContainer fc2 = FunctContainer([](double x){return (x * x - 1); });
I want to write FunctContainer
's ctors so that the lifetime of the function stored in fc1
is the one of f
and the lifetime in fc2
of the contained function is the lifetime of fc2
itself.
I have written something (see below) but I'm not really satisfied (I got it wrong).
This is correct c++ but wrong behavior: f_
expires after the call to the constructor when f_
is an rvalue.
class FunctContainerWRONG{
public:
IntegrandNotProjecting(const std::function<double(double)>& function)
: f_(function){}
IntegrandNotProjecting(const std::function<double(double)>&& function)
: f_(std::move(function)){}
const std::function<double(double)>& f_;
private:
};
This looks awful at me and probably is not correct c++ but is intended to explain in pseudocode what the desired behavior looks like. If possible I want to avoid to constuct a brand new object and I just want to let my object "persist":
class FunctContainer{
public:
FunctContainer(const std::function<double(double)>& function)
: f_p(nullptr),
f_(function){}
FunctContainer(const std::function<double(double)>&& function)
: f_p()std::make_shared<std::function<double(double)>>(function)),
f_(*f_p){}
private:
std::shared_ptr<std::function<double(double)>> f_p;
const std::function<double(double)>& f_;
};