Understanding operator overloading 'const' syntax for use in STL map - c++

I am trying to understand the syntax of operator overloading for operator < where the second const word needed for the compiler to be happy.
bool operator < (const point& rhs) const { // what's the rationale for the second const word here?
Take for example this struct
struct point {
int x;
int y;
point () : x(0), y(0) {}
point (int _x, int _y) : x(_x), y(_y) {}
point operator + (const point& rhs) {
return point(this->x + rhs.x, this->y + rhs.y);
}
bool operator < (const point& rhs) const {
return this->x < rhs.x;
}
};
This will allow me to use it as key to mymap.
map<point,int> mymap;

The outer const on the end of a method declaration tells the compiler that the method's *this object is to be const.
std::map stores its keys as const values. So any operators that are applied to the keys need to be marked as const, or else they will fail to compile. std::map uses operator< by default to order its keys and compare them for equality.
Besides, as good practice, any member method/operator that does not modify the contents of *this should be marked as const anyway. Doing so lets the user know that such operations are meant to be read-only, and lets the compile use them in expressions on const objects.

The const at the end means that the function will not change the object it is invoked on. This allows the function to be invoked on const objects.

That second const is a qualifier for the object pointed by the implied argument this. It means the method is not allowed to modify its object. Indeed, comparison needn't - and shouldn't - modify the object being compared.

Related

The method find of a set of none const values cannot be called with const values. Can const_cast be used to solve that problem?

Let us say that p below has to be a pointer to const X. Then it is not possible to call find for a set of pointers to X with my special compare class. Is that a shortcoming of 'set' and 'find'? Is it safe to solve it with const_cast as I have done?
struct X{
std::string key;
X(std::string s): key(s) {}
};
struct compare {
bool operator() (const X* lhs, const X* rhs) const {
return lhs->key < rhs->key;
}
};
int main() {
std::set<X*,compare> m;
const X a("hello");
const X*p=&a;
std::set<X*,compare>::const_iterator it=m.find(const_cast<X*>(p));
}
This use of const_cast is safe, but any usage of const_cast is scary. const_cast is legal so long as you don't modify the object through the cast, which std::set::find does not do.
However, you don't need a const_cast here. If you make your comparator transparent, that opts into allowing find to search based on anything comparable to the key type. This is exactly what we want:
struct compare {
using is_transparent = void; // doesn't matter which type you use
bool operator() (const X* lhs, const X* rhs) const {
// You might want to consider using std::less<X*> to compare these.
// std::less<T*> can compare pointers which are not part of the
// same array, which is not allowed with just a simple less-than
// comparison.
return lhs->key < rhs->key;
}
};
Complete example: https://godbolt.org/z/NsZccs

Invalid Operands to binary expression(const point and const point

map<pair<int,int>,int>pairOfNumbers;
pairOfNumbers.insert(pair<pair<int,int>,int>({1,2},2));
this is working, but
map<pair<point,point>,int>PointsOnLine;
PointsOnLine.insert(pair<pair<point,point>,int>(make_pair(points[i],points[j]),count));
this doesn't.
point is just a structure of two ints x and y;
I keep getting the error 'Invalid Operands to binary expression(const point and const point' this is the structure of point.
struct point
{
int x;
int y;
public:
bool operator==(const point& p)
{
if(x==p.x && y==p.y)
return true;
else
return false;
}
bool operator!=(const point& p)
{
if(x==p.x &&y==p.y)
return false;
else
return true;
}
};
how do I insert two points and distance between them in the map?
in Xcode I get this error
Your point type does not support weak-ordering. It has no method of determining is-less-than. You may think you don't need that because your point is actually tucked into a std::pair<point,point> but you do.
std::pair<T1,T2> supports weak ordering only if T1 and T2 do. In your case, they're the same type, so for a std::pair<point,point> to be used as key in a std::map<std::pair<point,pint>,T>, point must support weak ordering.
To support weak ordering, you must either provide an operator< that compares two of your objects in question, or a comparator functor type that does the same thing. The easiest way for you to do this would be:
#include <tuple>
struct point
{
int x;
int y;
bool operator <(const point& p) const
{
return std::tie(x, y) < std::tie(p.x, p.y);
}
bool operator ==(const point& p) const
{
return !(*this < p || p < *this);
}
bool operator !=(const point& p) const
{
return *this < p || p < *this;
}
};
I took liberty to redefine operator == and operator != to utilize the weak order properties of the proper operator <. It wasn't necessary, but ultimately it's just easier if operators root to as basic code as possible. With the above change you should be able to use both point and std::pair<point,point> as key types in std::map and std::set. In truth, a strict weak ordering can define all of the basic comparators (<=, >, >=, !=, ==) as derivations from operator < in one form or another (or relative to something that does). I challenge you to consider them, try implementing them, and above all, writing some test harnesses that verify your implementation.

C2678 when compiling. Visual Studio refers to xutility [duplicate]

This question already has answers here:
How to use std::find() with vector of custom class?
(2 answers)
Closed last year.
If I have a class
class Point
{
public:
Point() {}
Point(int _col, int _row) : row(_row), col(_col) {}
int row, col;
};
how can I use std::find() to check whether the point is already in vector? DO I have to overload operator== ?
I am trying to do this
#include <algorithm>
if(std::find(v.begin(), v.end(), x) != v.end()) {
/* v contains x */
} else {
/* v does not contain x */
}
Almost every answer I find on Stack Overflow suggest using find to check whether the object is in std::vector but none of them explains whether it compares the pointer of objects or the actual values of the object.
The C++ standard (draft N3242) says (in section 25.2.5 [alg.find]) that std::find:
Returns: The first iterator i in the range [first,last) for which the following corresponding conditions hold: *i == value[...]. Returns last if no such iterator is found.
Your question of whether it will search based on the value or the address of the object depends on how operator== is implemented. The simple answer is: std::find will return an iterator to the object for which operator== returned true.
Usually, this will just be a value-based comparison (because operator== is usually implemented to compare the values of two objects), and so you should generally expect std::find to search the range for the value you've provided (not the address of the object you provided).
It's possible for operator== to be implemented such that it compares based on address, like so:
bool operator==(const Point& left, const Point& right) {
return &left == &right;
}
Using this operator== will compare addresses, and so std::find will search for an object that has the same address as the one you've provided. It's generally a bad idea to implement operator== like this, though. Most people would implement operator== like so:
bool operator==(const Point& left, const Point& right) {
return left.row == right.row && left.col == right.col;
}
which, when used with std::find, will compare Points based on their values.
Unless your types are PODs fundamental types, you will need to provide an equality function, member or not.
There are two fundamental versions of std::find, one that assumes an equality operator and the other uses an equality function you supply.
I recommend that you add operator== and operator< to any class that will be compared for equality or ordered.
Here's an updated version of your class:
class Point
{
int x; // These are private by default.
int y;
public:
Point(int new_x, int new_y) : x(new_x), y(new_y)
{ ; }
bool operator==(const Point& p) const
{
return (x == p.x) && (y == p.y);
}
};
The member method operator== allows comparison without exposing the values to friends or the public.
If you want to use a free standing comparison function, you will need to either make the values public or make the function a friend:
class Point
{
int x; // These are private by default.
int y;
public:
Point(int new_x, int new_y) : x(new_x), y(new_y)
{ ; }
friend bool operator==(const Point& a, const Point& b);
};
bool operator==(const Point& a, const Point& b)
{
return (a.x == b.x) && (a.y == b.y);
}
If you want to use the free standing function with std::find, the example would be:
std::vector<Point> point_container;
//...
Point p;
std::vector<Point>::const_iterator iter;
iter = std::find(point_container.begin(), point_container.end(),
p,
Equal_Points);
Where Equal_Points is a free standing function that can compare the members of two Points.

std::find how does it work? operator== [duplicate]

This question already has answers here:
How to use std::find() with vector of custom class?
(2 answers)
Closed last year.
If I have a class
class Point
{
public:
Point() {}
Point(int _col, int _row) : row(_row), col(_col) {}
int row, col;
};
how can I use std::find() to check whether the point is already in vector? DO I have to overload operator== ?
I am trying to do this
#include <algorithm>
if(std::find(v.begin(), v.end(), x) != v.end()) {
/* v contains x */
} else {
/* v does not contain x */
}
Almost every answer I find on Stack Overflow suggest using find to check whether the object is in std::vector but none of them explains whether it compares the pointer of objects or the actual values of the object.
The C++ standard (draft N3242) says (in section 25.2.5 [alg.find]) that std::find:
Returns: The first iterator i in the range [first,last) for which the following corresponding conditions hold: *i == value[...]. Returns last if no such iterator is found.
Your question of whether it will search based on the value or the address of the object depends on how operator== is implemented. The simple answer is: std::find will return an iterator to the object for which operator== returned true.
Usually, this will just be a value-based comparison (because operator== is usually implemented to compare the values of two objects), and so you should generally expect std::find to search the range for the value you've provided (not the address of the object you provided).
It's possible for operator== to be implemented such that it compares based on address, like so:
bool operator==(const Point& left, const Point& right) {
return &left == &right;
}
Using this operator== will compare addresses, and so std::find will search for an object that has the same address as the one you've provided. It's generally a bad idea to implement operator== like this, though. Most people would implement operator== like so:
bool operator==(const Point& left, const Point& right) {
return left.row == right.row && left.col == right.col;
}
which, when used with std::find, will compare Points based on their values.
Unless your types are PODs fundamental types, you will need to provide an equality function, member or not.
There are two fundamental versions of std::find, one that assumes an equality operator and the other uses an equality function you supply.
I recommend that you add operator== and operator< to any class that will be compared for equality or ordered.
Here's an updated version of your class:
class Point
{
int x; // These are private by default.
int y;
public:
Point(int new_x, int new_y) : x(new_x), y(new_y)
{ ; }
bool operator==(const Point& p) const
{
return (x == p.x) && (y == p.y);
}
};
The member method operator== allows comparison without exposing the values to friends or the public.
If you want to use a free standing comparison function, you will need to either make the values public or make the function a friend:
class Point
{
int x; // These are private by default.
int y;
public:
Point(int new_x, int new_y) : x(new_x), y(new_y)
{ ; }
friend bool operator==(const Point& a, const Point& b);
};
bool operator==(const Point& a, const Point& b)
{
return (a.x == b.x) && (a.y == b.y);
}
If you want to use the free standing function with std::find, the example would be:
std::vector<Point> point_container;
//...
Point p;
std::vector<Point>::const_iterator iter;
iter = std::find(point_container.begin(), point_container.end(),
p,
Equal_Points);
Where Equal_Points is a free standing function that can compare the members of two Points.

compare function for upper_bound / lower_bound

I want to find the first item in a sorted vector that has a field less than some value x.
I need to supply a compare function that compares 'x' with the internal value in MyClass but I can't work out the function declaration.
Can't I simply overload '<' but how do I do this when the args are '&MyClass' and 'float' ?
float x;
std::vector< MyClass >::iterator last = std::upper_bound(myClass.begin(),myClass.end(),x);
What function did you pass to the sort algorithm? You should be able to use the same one for upper_bound and lower_bound.
The easiest way to make the comparison work is to create a dummy object with the key field set to your search value. Then the comparison will always be between like objects.
Edit: If for some reason you can't obtain a dummy object with the proper comparison value, then you can create a comparison functor. The functor can provide three overloads for operator() :
struct MyClassLessThan
{
bool operator() (const MyClass & left, const MyClass & right)
{
return left.key < right.key;
}
bool operator() (const MyClass & left, float right)
{
return left.key < right;
}
bool operator() (float left, const MyClass & right)
{
return left < right.key;
}
};
As you can see, that's the long way to go about it.
You can further improve Mark's solution by creating a static instance of MyClassLessThan in MyClass
class CMyClass
{
static struct _CompareFloatField
{
bool operator() (const MyClass & left, float right) //...
// ...
} CompareFloatField;
};
This way you can call lower_bound in the following way:
std::lower_bound(coll.begin(), coll.end(), target, CMyClass::CompareFloatField);
This makes it a bit more readable
Pass a lambda function to upper_bound
float x;
MyClass target;
target.x_ = x;
std::vector< MyClass >::iterator last =
std::upper_bound(myClass.begin(),myClass.end(),target,
[](const MyClass& a, const MyClass& b){return a.x_ < b.x_;});
I think what you need is std::bind2nd(std::less<MyClass>(), x). But, of course, the operator< must be defined for MyClass.
Edit: oh and I think you will need a constructor for MyClass that accepts only a float so that it can be implicitly converted. However, there might be a better way to do this.