Move semantics to avoid temporary object creation - c++

I am trying to do operations between large objects and I experiment with r-value references to avoid temporary object creations.
The experiment is the following code, but the result is not what I expected.
The code:
#include <iostream>
using namespace std;
struct A
{
A() = default;
A(const A& a) { cout << "copy ctor" << endl; }
A(A&& a) { cout << "move ctor" << endl; }
A &operator=(const A& a) { cout << "copy assign" << endl; return *this; }
A &operator=(A&& a) { cout << "move assign" << endl; return *this; }
A &operator*=(double s) { cout << "this = this *= s" << endl; return *this; }
A operator*(double s) const { cout << "A = const this * s" << endl; return *this; }
A &operator+=(const A &b) { cout << "this = this + const A&" << endl; return *this; }
A operator+(const A &b) const { cout << "A = const this + const A&" << endl; return *this; }
A &operator+(A &&b) const { cout << "A&& = const this + A&& --> "; return b += *this; }
};
A &operator+(A &&a, const A &b) { cout << "A&& = A&& + const A& --> "; return a += b; }
A &operator*(A &&a, double s) { cout << "A&& = A&& * s --> "; return a *= s; }
int main()
{
A a,b,c,d;
a = b + a * 4 + /*operator*(static_cast<A&&>(d), 2)*/ d * 2 + (A() + c) * 5;
return 0;
}
The output:
A&& = A&& + const A& --> this = this + const A& // A() + c
A = const this * s // (...) * 5
copy ctor // ???
A = const this * s // d * 2
copy ctor // ???
A = const this * s // a * 4
copy ctor // ???
A&& = const this + A&& --> this = this + const A& // (d*2) + (...)
A&& = const this + A&& --> this = this + const A& // (a*4) + (...)
A&& = const this + A&& --> this = this + const A& // b + (...)
copy assign // a = (...)
What I expect:
A&& = A&& + const A& --> this = this + const A& // A() + c
A&& = A&& * s --> this = this *= s // (...) * 5
A&& = A&& * s --> this = this *= s // (...) * 2 d is not used anymore, so I want to move semantics
A = const this * s // a * 4 a is not used anymore, but I want to keep semantics
A&& = A&& + const A& --> this = this + const A& // (d*2) + (...)
A&& = A&& + const A& --> this = this + const A& // (a*4) + (...)
A&& = A&& + const A& --> this = this + const A& // b + (...)
move assign // a = (...)

Here's a more correct version with fewer copies:
#include <iostream>
#include <utility>
using namespace std;
struct A
{
A() = default;
A(const A& a) { cout << "copy ctor" << endl; }
A(A&& a) { cout << "move ctor" << endl; }
A &operator=(const A& a) { cout << "copy assign" << endl; return *this; }
A &operator=(A&& a) { cout << "move assign" << endl; return *this; }
A &operator*=(double s) { cout << "this *= s" << endl; return *this; }
A &operator+=(const A &b) { cout << "this += const A&" << endl; return *this; }
};
A&& operator+(A &&a, const A &b)
{ cout << "A&& + const A&" << endl; a+=b; return std::move(a); }
A&& operator+(A &&a, A &&b)
{ cout << "A&& + A&&" << endl; a+=b; return std::move(a); }
// I assume commutativity
A&& operator+(const A &a, A &&b)
{ cout << "const A& + A&&" << endl; b+=a; return std::move(b); }
A operator+(const A &a, const A &b)
{ cout << "const A& + const A&" << endl; A r(a); r+=b; return r; }
A&& operator*(A &&a, double s)
{ cout << "A&& * s" << endl; a*=s; return std::move(a); }
A operator*(const A& a, double s)
{ cout << "const A& * s" << endl; A r(a); r*=s; return r; }
int main()
{
A a,b,c,d;
a = b + a * 4 + d * 2 + (A() + c) * 5;
return 0;
}
and here's the (annotated) output with ts being temporaries created:
expression level actual operations
---------------- -----------------
const A& * s t1 = a * 4
copy ctor create t1 = copy a
this *= s t1 *= 4
const A& + A&& b + t1
this += const A& t1 += b
const A& * s t2 = d * 2
copy ctor create t2 = copy d
this *= s t2 *= 2
A&& + A&& t1 + t2
this += const A& t1 += t2
A&& + const A& A() + c (note: A() is already a temporary)
this += const A& A() += c
A&& * s A'() * 5
this *= s A'() *= 5
A&& + A&& t1 + A''()
this += const A& t1 += A''()
move assign a = t1 a = t1
I don't think you can expect it any better than just two temporaries for the whole expression.
Concerning your commented-out code: try std::move(d) instead of plain d and you will safe the copy of d in the above output and reduces the number of temporaries to one. If you also add std::move(a), the whole expression is evaluated without a single temporary!
Also note that without std::move(d) and std::move(a), the compiler has no clue that it should/could move those objects, so any code which ends up moving them anyways is dangerous and plain wrong.
Update: I turned my ideas into a library, find it at GitHub. With this, your code becomes as simple as:
#include <iostream>
using namespace std;
#include <df/operators.hpp>
struct A : df::commutative_addable< A >, df::multipliable< A, double >
{
A() = default;
A(const A& a) { cout << "copy ctor" << endl; }
A(A&& a) { cout << "move ctor" << endl; }
A &operator=(const A& a) { cout << "copy assign" << endl; return *this; }
A &operator=(A&& a) { cout << "move assign" << endl; return *this; }
A &operator*=(double s) { cout << "this *= s" << endl; return *this; }
A &operator+=(const A &b) { cout << "this += const A&" << endl; return *this; }
};
while still being efficient and avoiding any unneccesary temporaries. Enjoy!

First of all A() + c returns by lvalue reference. That makes the expression itself an lvalue.
A function call is an lvalue if the result type is an lvalue reference type or an rvalue reference to function type, an xvalue if the result type is an rvalue reference to object type, and a prvalue otherwise.
An lvalue can't bind to an rvalue reference, so the member version of operator* is chosen. Your non-member functions should probably be returning by value:
A operator+(A &&a, const A &b) { cout << "A&& = A&& + const A& --> "; return a += b; }
A operator*(A &&a, double s) { cout << "A&& = A&& * s --> "; return a *= s; }
This causes the result to continue to be a prvalue expression referring to a temporary object.
Second, the copy constructor calls are caused by the member operators returning by value. This will cause a copy of the object. For example, when (...) * 5 returns, it will copy the value of *this out of the function:
A operator*(double s) const { cout << "A = const this * s" << endl; return *this; }

Your operators are implemented to return by value / by lvalue reference.
This results in chained operations accepting either object copy (hence the copy ctor) or lvalue reference.
E.g. b + a * 4 is equal to b.operator+(a.operator*(4)). The input to operator+ will be copy of the object.

Here is the signatures of your methods:
struct A
{
A() = default;
A(const A& a);
A(A&& a);
A &operator=(const A& a);
A &operator=(A&& a);
A &operator*=(double s);
A operator*(double s) const;
A &operator+=(const A &b);
A operator+(const A &b) const;
A &operator+(A &&b) const;
};
A &operator+(A &&a, const A &b);
A &operator*(A &&a, double s);
Problems show up right here. First, free operator+ should return an A&& that it is passed in, to avoid changing the rvalue reference into an lvalue. The same is true of A &A::operator+(A &&b) const; -- it should return an A&&.
Next, your free operators are chaining into the += operators. Here is a cute technique:
template<typename T>
A&&operator+(A &&a, T&&b){ return std::move(a+=std::forward<T>(b)); }
template<typename T>
A&&operator*(A &&a, T&&b){ return std::move(a*=std::forward<T>(b)); }
where we blinding forward our arguments to the += operation.
This can be made more robust, error-wise, with the auto return value technique:
template<typename T>
auto operator+(A &&a, T&&b)->declval(std::move(a+=std::forward<T>(b)))
{ return std::move(a+=std::forward<T>(b)); }
template<typename T>
auto operator*(A &&a, T&&b)->declval(std::move(a*=std::forward<T>(b)))
{ return std::move(a*=std::forward<T>(b)); }
which bumps errors up 1 step in the parsing stack using SFINAE. (Note that the && in T&& and A&& have completely different meanings -- T&&'s && is being used in a type deduction context, so T can bind to any reference type, while A&&'s && is not being used in a type deduction context, so it means A&& binds to an rvalue.).
What follows next is a far more heavily marked up version with some basic modifications for both correctness and efficiencies sake. I keep track of the history of each instance in the name field -- manipulations of this field are not "real", and its value represents the "calculation" required to create a given instance.
I assume move operations move this state.
#include <iostream>
#include <utility>
struct A;
A &operator+=(A& a, std::string op);
A&&operator+=(A&& a, std::string op);
struct recurse_nl {
int& count() {
static int v = 0;
return v;
}
recurse_nl(){if (++count()>1) std::cout << " --> "; else if (count()>2) std::cout << " --> [";}
~recurse_nl(){if (--count() == 0) std::cout <<"\n"; else if (count()>1) std::cout << "]"; }
};
struct A
{
std::string name;
A() = delete;
A(std::string n):name(n) { recurse_nl _; std::cout << "AUTO ctor{"<<name<<"}";};
A(const A& o):name(o.name+"_c&") { recurse_nl _; std::cout << "COPY ctor{"<<name<<"}(const&)"; }
A(A&& o):name(std::move(o.name)) { recurse_nl _; std::cout << "ctor{"<<name<<"}(&&)"; }
A(A& o):name(o.name+"_&") { recurse_nl _; std::cout << "COPY ctor{"<<name<<"}(&)"; }
A &operator=(const A& rhs) { recurse_nl _; std::cout << "COPY assign{"<<name<<"}={"<<rhs.name<<"}"; this->name = rhs.name; return *this; }
A &operator=(A&& rhs) { recurse_nl _; std::cout << "move assign{"<<name<<"}={"<<rhs.name<<"}"; this->name = std::move(rhs.name); return *this; }
A &operator*=(double d) { recurse_nl _; std::cout << "this{"<<name<<"} *= s{"<<d<<"}"; return (*this) += "(*#)"; }
A operator*(double d) const { recurse_nl _; std::cout << "A = const this{"<<name<<"} * s{"<<d<<"}"; A tmp(*this); return std::move(tmp*=d); }
A &operator+=(const A &rhs) { recurse_nl _; std::cout << "this{"<<name<<"} += const A&{"<<rhs.name<<"}"; return ((*this)+="(+=")+=rhs.name+")"; }
A operator+(const A &rhs) const { recurse_nl _; std::cout << "A = const this{"<<name<<"} + const A&{"<<rhs.name<<"}"; return std::move(A(*this)+="(+)"); }
A&& operator+(A &&rhs) const { recurse_nl _; std::cout << "A&& = const this{"<<name<<"} + A&&{"<<rhs.name<<"}"; return std::move(rhs += *this); }
~A() { recurse_nl _; std::cout << "dtor{"<<name<<"}"; }
};
A &operator+=(A& a, std::string op)
{ a.name+=op; return a; }
A&&operator+=(A&& a, std::string op)
{ a.name+=op; return std::move(a); }
template<typename T>
struct ref_type_of {
std::string value() const { return "value"; }
};
template<typename T>
struct ref_type_of<T&> {
std::string value() const { return "&"; }
};
template<typename T>
struct ref_type_of<T&&> {
std::string value() const { return "&&"; }
};
template<typename T>
struct ref_type_of<T const&&> {
std::string value() const { return " const&&"; }
};
template<typename T>
struct ref_type_of<T const&> {
std::string value() const { return " const&"; }
};
template<typename T>
std::string ref_type() { return ref_type_of<T>().value(); }
template<typename T>
A&& operator+(A &&a, T&& b) { recurse_nl _; std::cout << "A&&{"<<a.name<<"} = A&&{"<<a.name<<"} + T" << ref_type<T>(); return std::move(a += std::forward<T>(b)); }
template<typename T>
A&& operator*(A &&a, T&& b) { recurse_nl _; std::cout << "A&&{"<<a.name<<"} = A&&{"<<a.name<<"} * T" << ref_type<T>(); return std::move(a *= std::forward<T>(b)); }
void test1()
{
A a("a"),b("b"),c("c"),d("d");
a = b + a * 4 + d * 2 + (A("tmp") + c) * 5;
}
int main()
{
std::cout << "test1\n";
test1();
return 0;
}
I played with this on live work space and here is the output:
stdout:
test1
AUTO ctor{a}
AUTO ctor{b}
AUTO ctor{c}
AUTO ctor{d}
AUTO ctor{tmp}
A&&{tmp} = A&&{tmp} + T& --> this{tmp} += const A&{c}
A&&{tmp(+=c)} = A&&{tmp(+=c)} * Tvalue --> this{tmp(+=c)} *= s{5}
A = const this{d} * s{2} --> COPY ctor{d_c&}(const&) --> this{d_c&} *= s{2} --> ctor{d_c&(*#)}(&&) --> dtor{}
A = const this{a} * s{4} --> COPY ctor{a_c&}(const&) --> this{a_c&} *= s{4} --> ctor{a_c&(*#)}(&&) --> dtor{}
A&& = const this{b} + A&&{a_c&(*#)} --> this{a_c&(*#)} += const A&{b}
A&&{a_c&(*#)(+=b)} = A&&{a_c&(*#)(+=b)} + Tvalue --> this{a_c&(*#)(+=b)} += const A&{d_c&(*#)}
A&&{a_c&(*#)(+=b)(+=d_c&(*#))} = A&&{a_c&(*#)(+=b)(+=d_c&(*#))} + Tvalue --> this{a_c&(*#)(+=b)(+=d_c&(*#))} += const A&{tmp(+=c)(*#)}
move assign{a}={a_c&(*#)(+=b)(+=d_c&(*#))(+=tmp(+=c)(*#))}
dtor{a}
dtor{d_c&(*#)}
dtor{tmp(+=c)(*#)}
dtor{d}
dtor{c}
dtor{b}
dtor{a_c&(*#)(+=b)(+=d_c&(*#))(+=tmp(+=c)(*#))}
which is pretty verbose, but demonstrates pretty much every operation.
I modified your code so that operator+ and operator* actually creates a new object when required. The expensive operations (creating a new object, and copying) are highlighted through use of AUTO and COPY -- as you can see, there are the initial 4 alphabet objects, the tmp object in the expression, and two copies created by operator*(double).
We can get rid of some of the copies with this:
a = b + std::move(a) * 4 + std::move(d) * 2 + (A("tmp") + c) * 5;
however, we still end up with 3 objects with non-trivial state to destroy because twice we do operator+(A&&, A&&), and I didn't assume that this operation is extra efficient.
If it is, we can add this operator:
A &operator+=(A &&rhs) { recurse_nl _; std::cout << "this{"<<name<<"} += A&&{"<<rhs.name<<"}"; return ((*this)+="(+=")+=std::move(rhs.name)+")"; }
and the resulting output shows that only one object with non-trivial state is ever destroyed.
Final version on live workspace is here.
(The recurse_nl object is for recursion tracking. At the base level, it prints a newline at the end of the function. At deeper recursions, it does the --> printing, and in theory if recursion got deep enough it would print [ brackets to help).
Final output:
test1
AUTO ctor{a}
AUTO ctor{b}
AUTO ctor{c}
AUTO ctor{d}
AUTO ctor{tmp}
A&&{tmp} = A&&{tmp} + T& --> this{tmp} += const A&{c}
A&&{tmp(+=c)} = A&&{tmp(+=c)} * Tvalue --> this{tmp(+=c)} *= s{5}
A&&{d} = A&&{d} * Tvalue --> this{d} *= s{2}
A&&{a} = A&&{a} * Tvalue --> this{a} *= s{4}
A&& = const this{b} + A&&{a(*#)} --> this{a(*#)} += const A&{b}
A&&{a(*#)(+=b)} = A&&{a(*#)(+=b)} + Tvalue --> this{a(*#)(+=b)} += A&&{d(*#)}
A&&{a(*#)(+=b)(+=d(*#))} = A&&{a(*#)(+=b)(+=d(*#))} + Tvalue --> this{a(*#)(+=b)(+=d(*#))} += A&&{tmp(+=c)(*#)}
move assign{a(*#)(+=b)(+=d(*#))(+=tmp(+=c)(*#))}={a(*#)(+=b)(+=d(*#))(+=tmp(+=c)(*#))}
dtor{}
dtor{}
dtor{c}
dtor{b}
dtor{a(*#)(+=b)(+=d(*#))(+=tmp(+=c)(*#))}
where you can see the single "complex object" destroyed at the end (together with its entire history).

Related

c++ multiple operator overloading

I want to overload some operators multiple times and i am not sure how to do it i have a class
for example
int main(){
addClass c1;
addClass c2;
addClass c1 = (c2 + 10 * c2 + 2 * c3)*c3;
}
my problem is that i dont know how to make this many opreator ovrloading, I mean what do i need to return in ordear for it to continue to the next opreator?
I tried that
class addClass
//declration
friend addClass & operator*(const double, const addClass &);
addClass & operator*(double x, const addClass & a1)
{
int i;
addClass add(a1.rank);
add.res=x*a1.res
}
return res;
}
but for the next oprators i run into some problems so i wanted to know what to do
Thank you!
The order of evaluation is the same as usual even for overloaded operators, so with this example where I've added printing in the operators ...
#include <iostream>
class addClass {
public:
addClass() : addClass(0) {}
// converting constructor
addClass(int x) : value(x) {}
addClass& operator+=(const addClass& rhs) {
value += rhs.value;
return *this;
}
addClass& operator-=(const addClass& rhs) {
value -= rhs.value;
return *this;
}
addClass& operator*=(const addClass& rhs) {
value *= rhs.value;
return *this;
}
addClass& operator/=(const addClass& rhs) {
value /= rhs.value;
return *this;
}
int getValue() const { return value; }
private:
int value;
};
std::ostream& operator<<(std::ostream& os, const addClass& ac) {
return os << ac.getValue();
}
addClass operator+(addClass lhs, const addClass& rhs) {
std::cout << lhs << '+' << rhs << '=';
lhs += rhs;
std::cout << lhs << '\n';
return lhs;
}
addClass operator-(addClass lhs, const addClass& rhs) {
std::cout << lhs << '-' << rhs << '=';
lhs -= rhs;
std::cout << lhs << '\n';
return lhs;
}
addClass operator*(addClass lhs, const addClass& rhs) {
std::cout << lhs << '*' << rhs << '=';
lhs *= rhs;
std::cout << lhs << '\n';
return lhs;
}
addClass operator/(addClass lhs, const addClass& rhs) {
std::cout << lhs << '/' << rhs << '=';
lhs /= rhs;
std::cout << lhs << '\n';
return lhs;
}
int main(){
addClass c1;
addClass c2 = 100;
addClass c3 = 2;
c1 = (c2 + 10 * c2 / 2 * c3)*c3;
}
... you would get this output:
10*100=1000
1000/2=500
500*2=1000
100+1000=1100
1100*2=2200

How to have inheritance between template with union?

I have the following two objects. Im wondering if there is a way to have Pixel as a base class of PixelBGR so that any operator (+,-,*,/, [], etc.) could be used without redefining them ?
template<class T, std::size_t N>
struct Pixel
{
T ch[N];
inline T& operator[](const int x)
{
return ch[x];
}
};
template<class T>
struct PixelBGR
{
union
{
struct
{
T b;
T g;
T r;
};
T ch[3];
};
inline T& operator[](const int x)
{
return ch[x];
}
};
EDIT: As suggested by πάντα ῥεῖ, here more details about what Im trying to do.
Im trying to have a generic class Pixel, which will be template to handle any type or size.
The usual are 1,2,3,4,8 or 16. The class with defines some operator such as +,-,*, etc.
Since most of the time, the Pixel<T,3> is a BGR pixel, I would like to define rapid access to r,g and b to avoid confusion, but still store it as BGR.
But the derived class should also provide the Operator which will be generic based on N.
EDIT2: By reading the comment of SergeyA, I forgot to say that the struct Pixel must not change size.
So I think balki answer is the best, by using member function. I was trying to make it with variables to avoid too much char ie: adding the (), but it seems to be too complicated for nothing. I still investigating CRTP, but I dont get it well, Im reading on that.
Answering the question as asked, this should give OP reuse of the operators without any undefined behavior:
#include <cstddef>
template<class T, std::size_t N>
struct Pixel
{
T ch[N];
inline T& operator[](const int x)
{
return ch[x];
}
Pixel& operator+= (const Pixel& ) { return *this;}
};
template<class T, std::size_t N>
Pixel<T, N> operator+ (const Pixel<T, N>& l, const Pixel<T, N>& r);
template<class T>
struct BgrPixel : Pixel<T, 3> {
using base = Pixel<T, 3>;
using base::base;
BgrPixel(const base& b) : base(b) { };
T& b = base::ch[0];
T& g = base::ch[1];
T& r = base::ch[2];
};
BgrPixel<int> a, b;
BgrPixel<int> c = a + b;
Alterinative would be to have b(), g() and r() as a member functions, but this would require you to access them as functions. You would also need const and non-const versions of them.
The benefits, however, would be that the size of the struct will not be increased and copy assignment would work naturally (which, in turn, could be solved by providing custom copy assignment).
The Curiously Recurring Template Pattern (CRTP) would work well in this case. In the CRTP the Derived class is used as a template argument to the Base class. Chapter 16.3 The Curiously Recurring Template Pattern (CRTP), from the C++ Templates - The Complete Guide, by David Vandevoorde and Nicolai M. Josuttis, explains things in more detail.
From the comments below, the usage of a union{struct{...}...} causes undefined behaviour (UB), but there have been some contradicting opinions upon this. As far as I'm aware, it is a gnu extension and supported by almost every compiler. glm for example uses union-structs quite very often.
As an alternative approach, you can use aliases (references) for the r,g,b variables.
#include <iostream>
template<typename T, std::size_t N, template<typename,std::size_t> class B >
struct Pixel
{
B<T,N> *crtp = static_cast<B<T,N>*>(this);
T& operator[](std::size_t x)
{
return crtp->ch[x];
}
Pixel& operator = (const Pixel &t)
{
crtp->ch[0] = t.crtp->ch[0];
crtp->ch[1] = t.crtp->ch[1];
crtp->ch[2] = t.crtp->ch[2];
return *crtp;
}
B<T,N> operator + (const B<T,N> &t)
{
B<T,N> tmp;
tmp[0] = crtp->ch[0] + t.crtp->ch[0];
tmp[1] = crtp->ch[1] + t.crtp->ch[1];
tmp[2] = crtp->ch[2] + t.crtp->ch[2];
return tmp;
}
B<T,N> operator - (const B<T,N> &t)
{
B<T,N> tmp;
tmp[0] = crtp->ch[0] - t.crtp->ch[0];
tmp[1] = crtp->ch[1] - t.crtp->ch[1];
tmp[2] = crtp->ch[2] - t.crtp->ch[2];
return tmp;
}
};
template<typename T, std::size_t N=3>
struct PixelBGR : Pixel<T, N, PixelBGR>
{
T ch[3];
T &r;
T &g;
T &b;
PixelBGR() : ch{},r(ch[0]),g(ch[1]),b(ch[2])
{}
PixelBGR& operator = (const PixelBGR &p)
{
ch[0] = p.ch[0];
ch[1] = p.ch[1];
ch[2] = p.ch[2];
return *this;
}
};
int main()
{
PixelBGR<int> p;
p.r = 25;
p.g = 14;
p.b = 58;
std::cout<< p[0] <<" , "<<p[1]<<" , "<<p[2] <<std::endl;
PixelBGR<int> q;
q = p;
std::cout<< q[0] <<" , "<<q[1]<<" , "<<q[2] <<std::endl;
PixelBGR<int> res1;
res1 = q + p;
std::cout<< res1.r <<" , "<<res1.g<<" , "<<res1.b <<std::endl;
PixelBGR<int> res2;
res2 = q - p;
std::cout<< res2.r <<" , "<<res2.g<<" , "<<res2.b <<std::endl;
}
Result:
25 , 14 , 58
25 , 14 , 58
50 , 28 , 116
0 , 0 , 0
Example using references: https://rextester.com/AZWG4319
Example using union-struct: https://rextester.com/EACC87146
First thanks to all of you for advise, and special thanks to #Constantinos Glynos, #balki and #SergeyA for the example they provide, those help me to achieve a solution that match my need.
I implemented the BGR and BGRA to show that the N works fine, now I just need to implement all the operator, and functions that I require.
Please feel free to edit, or tell me if there is something wrong with this.
Pixel.h
#include <cstdio> // std::size_t
#include <iostream> // std::cout
template<typename T, std::size_t N, template<typename, std::size_t> class B >
struct Pixel
{
T ch[N];
// ==============================================================
// Overload the accessor (so .ch[0] == direct access with [0].
T& operator[](std::size_t x){ return ch[x]; }
// ==============================================================
// Copy-assignement
Pixel& operator=( const Pixel &t )
{
for ( int i = 0; i < N; i++ )
ch[i] = t.ch[i];
return *this;
}
// ==============================================================
// Operator
B<T, N> operator+( const B<T, N> &t )
{
B<T, N> tmp;
for ( int i = 0; i < N; i++ )
tmp[i] = ch[i] + t.ch[i];
return tmp;
}
B<T, N> operator-( const B<T, N> &t )
{
B<T, N> tmp;
for ( int i = 0; i < N; i++ )
tmp[i] = ch[i] - t.ch[i];
return tmp;
}
template<typename T, std::size_t N, template<typename, std::size_t> class B >
friend std::ostream& operator<<( std::ostream& os, const Pixel &t );
};
// To print the vector
template<typename T, std::size_t N, template<typename, std::size_t> class B >
std::ostream& operator<<( std::ostream& os, const B<T, N> &t )
{
os << "Pixel: (" << t.ch[0];
for ( int i = 1; i < N; i++ )
os << ", " << t.ch[i];
os << ")";
return os;
}
template<typename T, std::size_t N = 3>
struct BGR : Pixel<T, N, BGR>
{
T& b() { return ch[0]; }
T& g() { return ch[1]; }
T& r() { return ch[2]; }
};
template<typename T, std::size_t N = 4>
struct BGRA : Pixel<T, N, BGRA>
{
T& b() { return ch[0]; }
T& g() { return ch[1]; }
T& r() { return ch[2]; }
T& a() { return ch[3]; }
};
Main.cpp
int main() {
std::cout << "Sizeof a float BGR: " << sizeof(BGR<float>) << std::endl;
std::cout << "Sizeof a float BGRA: " << sizeof(BGRA<float>) << std::endl;
BGR<int> p;
p.r() = 25;
p.g() = 14;
p.b() = 58;
std::cout << p << std::endl;
std::cout << p[0] << " , " << p[1] << " , " << p[2] << std::endl;
std::cout << p.b() << " , " << p.g() << " , " << p.r() << std::endl;
BGR<int> q;
q = p;
std::cout << q[0] << " , " << q[1] << " , " << q[2] << std::endl;
BGR<int> res1;
res1 = q + p;
std::cout << res1.r() << " , " << res1.g() << " , " << res1.b() << std::endl;
BGR<int> res2;
res2 = q - p;
std::cout << res2.r() << " , " << res2.g() << " , " << res2.b() << std::endl;
BGRA<float> a;
a.r() = 255.0f;
a.g() = 0.0f;
a.b() = 0.0f;
a.a() = 128.5f;
BGRA<float> b = a;
std::cout << a << std::endl;
return 0;
}

Can someone explain why it's fix the error? invalid initialization of non-const reference of type 'B&' from an rvalue of type 'B'?

Given the following code:
#include <iostream>
using namespace std;
class B {
private:
int n;
public:
B(int x) :
n(x) {
}
B operator+(B& b) {
return B(n + b.n);
}
friend ostream& operator<<(ostream &out, const B& b) {
out << "B: " << b.n;
return out;
}
bool operator<(const B& rhs) const {
return n < rhs.n;
}
};
int main() {
B b1(1), b2(2), b3(3);
const B b4 = b1 + (b2 + b3); // error
cout << b4 << b1 << b2 << b3;
return 0;
}
I know that if I will change this :
B operator+(B& b)
To:
B operator+(const B& b)
So it's will be ok. But, why it's really fix it?
invalid initialization of non-const reference of type 'B&' from an rvalue of type 'B'
(I understand the meaning of the error, but, I don't understand why the changing fix the error).
In addition, what is the difference between the code above (that give me error), to the following code:
class D: public B {
double m;
public:
D() :
B(0), m(0) {
}
D(int x, int y) :
B(x), m(y) {
}
D operator+(D& d) { // ~~
return d;
}
friend ostream& operator<<(ostream &out, const D& d) {
out << static_cast<const B&>(d);
out << ", D: " << d.m;
return out;
}
};
int main() {
B b4(4);
D d1(1, 0), d2(1, 2);
D d3 = d1 + d2;
cout << b4 << ", " << d3 << endl;
return 0;
}
Why now I don't get at ~~ also error?

How to avoid forgetting to define operator== in subclasses?

I have a base class and define a operator== on it. And B is a subclass of A and I forget to define operator== on B. Then A::operator== is used on comparing B and usually this gives an unexpected results. Any good method to avoid such "forget"? I add an example to clarify my question.
class A
{
public:
bool operator==(const A& rhs) const
{
return i == rhs.i;
}
int i
};
class B : public A
{
public:
int j;
}
B b1, b2;
b1.i = 1; b1.j = 2;
b2.i = 1; b1.j = 3;
bool b = (b1 == b2); // will be true
What you could try is put A in a namespace, create operator == as a template non-member also in that namespace and let ADL take care of it.
#include <iostream>
namespace stuff {
class A
{
};
class B : public A {};
template <typename T>
bool operator == (const T &lhs, const T &rhs)
{
std::cout << __PRETTY_FUNCTION__ << '\n';
return &lhs == &rhs; // <-- replace this with something real
}
}
struct C {};
int main()
{
stuff::A a, aa;
stuff::B b, bb;
C c, cc;
b == bb;
aa == a;
aa == cc; // error: no match for "operator==" stuff::A and C
b == a; // error: no match for "operator==" stuff::B and stuff::A
}
Edit: For your edited example where you want the equality check to compare each part of the class with the other respective corresponding part, DyP's suggestion can work. For example:
// same as before
// ...
class A
{
public:
bool is_equal(const A &rhs) const { return i == rhs.i; }
};
class B : public A
{
public:
bool is_equal(const B &rhs) const { return A::is_equal(rhs) && (j == rhs.j); }
};
template <typename T>
bool operator == (const T &lhs, const T &rhs)
{
std::cout << __PRETTY_FUNCTION__ << '\n';
return lhs.is_equal(rhs);
}
Now comparing this again in the using code:
// ...
b.i = 1, bb.i = 1;
b.j = 1, bb.j = 42;
cout << boolalpha << (b == bb) << '\n';
b.j = 42;
cout << (b == bb) << '\n';
a.i = 2, aa.i = 3;
cout << (aa == a) << '\n';
outputs:
bool stuff::operator==(const T&, const T&) [with T = stuff::B]
false
bool stuff::operator==(const T&, const T&) [with T = stuff::B]
true
bool stuff::operator==(const T&, const T&) [with T = stuff::A]
false
Allowing implicit conversions for greatwolf's great approach is a bit tricky:
#include <type_traits>
namespace stuff
{
template<class T, class U>
bool operator== (const T &lhs, const U &rhs)
{
using namespace std;
static_assert(is_convertible<T, U>{} || is_convertible<U, T>{},
"invalid argument type");
static_assert
(
is_same<T, U>{}
|| ( not is_base_of<T, U>{} && not is_base_of<U, T>{})
, "use explicit casts to compare derived to base class types"
);
return is_equal(lhs, rhs);
}
template<class T>
bool is_equal(T const&, T const&)
{
// force compile-time failure when instantiating
static_assert(std::is_same<T, void>{},
"no free is_equal function for these argument types available");
return false;
}
class A
{
private:
int i;
friend bool is_equal(A const& lhs, A const& rhs)
{ return lhs.i == rhs.i; }
public:
A(int p_i) : i(p_i) {}
};
class B : public A
{
int j;
public:
B(int p_i, int p_j) : A(p_i), j(p_j) {}
};
class C : public A
{
private:
int j;
friend bool is_equal(C const& lhs, C const& rhs)
{
return is_equal(static_cast<A const&>(rhs),
static_cast<A const&>(lhs))
&& lhs.j == rhs.j;
}
public:
C(int p_i, int p_j) : A(p_i), j(p_j) {}
};
}
struct D
{
operator stuff::C() const
{
return stuff::C(1, 42);
}
};
#include <iostream>
int main()
{
stuff::A a(1), aa(1);
stuff::B b(1, 42), bb(1, 42);
stuff::C c(1, 42), cc(1, 42);
D d;
// commented lines invoke compilation failures
std::cout << "a == aa: " << (a == aa) << std::endl;
//std::cout << "a == b : " << (a == b ) << std::endl;
//std::cout << "b == bb: " << (b == bb) << std::endl;
//std::cout << "a == c : " << (a == c ) << std::endl;
std::cout << "c == cc: " << (c == cc) << std::endl;
std::cout << "d == c : " << (d == c ) << std::endl;
}
Why do you have equality comparison in a class hierarchy? In many cases, this indicates a problem with the design, with classes that don't properly behave like value types, but not properly like objects from a hierarchy either.

Having Public properties in c++ class

How do I have properties in C++ class, as you have in a C# class.
I don't want to have getter and setter methods.
You can use a solution similar to that Jon suggested, yet retaining ordinary C++ semantics using operator overloading. I've slightly modified Jon's code as following (explanations follow the code):
#include <iostream>
template<typename T>
class Accessor {
public:
explicit Accessor(const T& data) : value(data) {}
Accessor& operator=(const T& data) { value = data; return *this; }
Accessor& operator=(const Accessor& other) { this->value = other.value; return *this; }
operator T() const { return value; }
operator T&() { return value; }
private:
Accessor(const Accessor&);
T value;
};
struct Point {
Point(int a = 0, int b = 0) : x(a), y(b) {}
Accessor<int> x;
Accessor<int> y;
};
int main() {
Point p;
p.x = 10;
p.y = 20;
p.x++;
std::cout << p.x << "," << p.y << std::endl;
p.x = p.y = 15;
std::cout << p.x << "," << p.y << std::endl;
return 0;
}
We overload operator= to retain the usual assignment syntax instead of a function-call-like syntax. We use the cast operator as a "getter". We need the second version of the operator= to allow assignment of the second kind in main().
Now you can add to Accessor's constructor function pointers, or better - functors - to call as getters/setters in any way seems right to you. The following example assumes the setter function return bool to convey agreement to setting the new value, and the getter can just modify it on it's way out:
#include <iostream>
#include <functional>
#include <cmath>
template<typename T>
class MySetter {
public:
bool operator()(const T& data)
{
return (data <= 20 ? true : false);
}
};
template<typename T>
class MyGetter {
public:
T operator()(const T& data)
{
return round(data, 2);
}
private:
double cint(double x) {
double dummy;
if (modf(x,&dummy) >= 0.5) {
return (x >= 0 ? ceil(x) : floor(x));
} else {
return (x < 0 ? ceil(x) : floor(x));
}
}
double round(double r, int places) {
double off = pow(10.0L, places);
return cint(r*off)/off;
}
};
template<typename T, typename G = MyGetter<T>, typename S = MySetter<T>>
class Accessor {
public:
explicit Accessor(const T& data, const G& g = G(), const S& s = S()) : value(data), getter(g), setter(s) {}
Accessor& operator=(const T& data) { if (setter(data)) value = data; return *this; }
Accessor& operator=(const Accessor& other) { if (setter(other.value)) this->value = other.value; return *this; }
operator T() const { value = getter(value); return value;}
operator T&() { value = getter(value); return value; }
private:
Accessor(const Accessor&);
T value;
G getter;
S setter;
};
struct Point {
Point(double a = 0, double b = 0) : x(a), y(b) {}
Accessor<double> x;
Accessor<double> y;
};
int main() {
Point p;
p.x = 10.712;
p.y = 20.3456;
p.x+=1;
std::cout << p.x << "," << p.y << std::endl;
p.x = p.y = 15.6426;
std::cout << p.x << "," << p.y << std::endl;
p.x = p.y = 25.85426;
std::cout << p.x << "," << p.y << std::endl;
p.x = p.y = 19.8425;
p.y+=1;
std::cout << p.x << "," << p.y << std::endl;
return 0;
}
However, as the last line demonstrates it has a bug. The cast operator returning a T& allows users to bypass the setter, since it gives them access to the private value. One way to solve this bug is to implement all the operators you want your Accessor to provide. For example, in the following code I used the += operator, and since I removed the cast operator returning reference I had to implement a operator+=:
#include <iostream>
#include <functional>
#include <cmath>
template<typename T>
class MySetter {
public:
bool operator()(const T& data) const {
return (data <= 20 ? true : false);
}
};
template<typename T>
class MyGetter {
public:
T operator() (const T& data) const {
return round(data, 2);
}
private:
double cint(double x) const {
double dummy;
if (modf(x,&dummy) >= 0.5) {
return (x >= 0 ? ceil(x) : floor(x));
} else {
return (x < 0 ? ceil(x) : floor(x));
}
}
double round(double r, int places) const {
double off = pow(10.0L, places);
return cint(r*off)/off;
}
};
template<typename T, typename G = MyGetter<T>, typename S = MySetter<T>>
class Accessor {
private:
public:
explicit Accessor(const T& data, const G& g = G(), const S& s = S()) : value(data), getter(g), setter(s) {}
Accessor& operator=(const T& data) { if (setter(data)) value = data; return *this; }
Accessor& operator=(const Accessor& other) { if (setter(other.value)) this->value = other.value; return *this; }
operator T() const { return getter(value);}
Accessor& operator+=(const T& data) { if (setter(value+data)) value += data; return *this; }
private:
Accessor(const Accessor&);
T value;
G getter;
S setter;
};
struct Point {
Point(double a = 0, double b = 0) : x(a), y(b) {}
Accessor<double> x;
Accessor<double> y;
};
int main() {
Point p;
p.x = 10.712;
p.y = 20.3456;
p.x+=1;
std::cout << p.x << "," << p.y << std::endl;
p.x = p.y = 15.6426;
std::cout << p.x << "," << p.y << std::endl;
p.x = p.y = 25.85426;
std::cout << p.x << "," << p.y << std::endl;
p.x = p.y = 19.8425;
p.y+=1;
std::cout << p.x << "," << p.y << std::endl;
return 0;
}
You'll have to implements all the operators you're going to use.
For behaviour that's kind of like this, I use a templated meta-accessor. Here's a highly simplified one for POD types:
template<class T>
struct accessor {
explicit accessor(const T& data) : value(data) {}
T operator()() const { return value; }
T& operator()() { return value; }
void operator()(const T& data) { value = data; }
private:
accessor(const accessor&);
accessor& operator=(const accessor&);
T value;
};
Typical usage is like this:
struct point {
point(int a = 0, int b = 0) : x(a), y(b) {}
accessor<int> x;
accessor<int> y;
};
point p;
p.x(10);
p.y(20);
p.x()++;
std::cout << p.x();
The compiler typically inlines these calls if you set things up right and have optimisation turned on. It's no more of a performance bottleneck than using actual getters and setters, no matter what optimisations happen. It is trivial to extend this to automatically support non-POD or enumerated types, or to allow callbacks to be registered for whenever data are read or written.
Edit: If you prefer not to use the parentheses, you can always define operator=() and an implicit cast operator. Here's a version that does just that, while also adding basic "stuff happened" callback support:
Further Edit: Okay, totally missed that someone already made a revised version of my code. Sigh.
If you don't care that your C++ code won't compile with anything other than the Microsoft Visual C++ compiler, then you can use some of the compiler's non-standard extensions.
For instance, the following code will create a C#-like property called MyProperty.
struct MyType
{
// This function pair may be private (for clean encapsulation)
int get_number() const { return m_number; }
void set_number(int number) { m_number = number; }
__declspec(property(get=get_number, put=set_number)) int MyProperty;
private:
int m_number:
}
int main()
{
MyType m;
m.MyProperty = 100;
return m.MyProperty;
}
More information on this Microsoft-specific language extension is available here.
Here's a PoC implementation I did a while back, works nicely except that you need to set something up in the constructor for it to work nice and smoothly.
http://www.codef00.com/code/Property.h
Here's the example usage:
#include <iostream>
#include "Property.h"
class TestClass {
public:
// make sure to initialize the properties with pointers to the object
// which owns the property
TestClass() : m_Prop1(0), m_Prop3(0.5), prop1(this), prop2(this), prop3(this) {
}
private:
int getProp1() const {
return m_Prop1;
}
void setProp1(int value) {
m_Prop1 = value;
}
int getProp2() const {
return 1234;
}
void setProp3(double value) {
m_Prop3 = value;
}
int m_Prop1;
double m_Prop3;
public:
PropertyRW<int, TestClass, &TestClass::getProp1, &TestClass::setProp1> prop1;
PropertyRO<int, TestClass, &TestClass::getProp2> prop2;
PropertyWO<double, TestClass, &TestClass::setProp3> prop3;
};
and some usage of this class...
int main() {
unsigned int a;
TestClass t;
t.prop1 = 10;
a = t.prop1;
t.prop3 = 5;
a = t.prop2;
std::cout << a << std::endl;
return 0;
}
There are two annoyances with this approach:
You need to give the property a
pointer to its owning class.
The syntax to declare a property is a
bit verbose, but I bet I can clean
that up a bit with some macros
You don't. C++ doesn't support properties like C# does. If you want code to run on set/get, it will have to be a method.
Properties aren't supported in C++, but you can implement them:
1) By using templates
2) By making language extension and writing custom code preprocessor
Either approach won't be easy, but it can be done.
You could provide get and set methods that have similar names to the data members:
class Example
{
private:
unsigned int x_;
double d_;
std::string s_s;
public:
unsigned int x(void) const
{ return x_;}
void x(unsigned int new_value)
{ x_ = new_value;}
double d(void) const
{ return d_;}
void d(double new_value)
{ d_ = new_value;}
const std::string& s(void) const
{ return s_;}
void s(const std::string& new_value)
{ s_ = new_value;}
};
Although this comes close, as it requires using '()' for each member, it doesn't meet the exact functionality of properties that Microsoft Languages provide.
The closest match for properties is to declare the data members as public.
Here’s a bit crude and simple implementation using a preprocessor macro to effortlessly generate nested classes that provide the functionality of getters and setters with the nice and clean syntax as if they were actual variables. No templates or function pointers are used (if that’s a plus), although your compiled program will have as many (sub)classes of the name property_* as there are PROPERTIES declarations, and like in Evan Teran’s solution, you need to give the property a reference to the owning class in the constructor.
More operators (operator++, operator+=) should be easy to add.
(I guess the reference to the owning class is ok, despite the circularity...? (x.property_age.x.property_age ...))
#include <iostream>
#include <stdexcept>
#define PROPERTY(type, name, owner, get_exprs, set_exprs) \
friend class property_ ##name; \
class property_ ##name { \
owner & x; \
public: \
property_ ##name (owner & init): x(init) {} \
operator type () { \
get_exprs \
} \
type operator= (const type value) { \
set_exprs \
return value; \
} \
} name ;
int current_year = 2020;
class person {
int year_of_birth; // Integer is signed to demonstrate argument validation
public:
// Remember to add each property and *this for them
person(): year_of_birth(0), age(*this) {}
const int& born() { return year_of_birth; }
// Remember the semicolons
PROPERTY(int, age, person,
/*get:*/ return current_year - x.year_of_birth; ,
/*set:*/ if (value < 0) throw std::invalid_argument("person::age : age cannot be negative");
x.year_of_birth = current_year - value; )
};
int main() {
person alice, bob;
alice.age = bob.age = 28;
alice.age = alice.age + 5;
//alice.age = -7; //throws
// Apparently the compiler does nice implicit conversion from
// (the macro-generated class) 'property_age' to 'int'
std::cout << "Alice: born: " << alice.born() << ", age: " << alice.age << '\n'
<< "Bob: born: " << bob.born() << ", age: " << bob.age << '\n'
<< "The mean of their ages is: " << (alice.age + bob.age) / 2.0 << '\n';
return 0;
}