Reputation: 413
This is a class that contains a const std::vector
and stores its min and max:
#include <vector>
#include <algorithm>
class MyClass {
public:
const std::vector<int> v;
const std::pair<int, int> minmax_v;
MyClass(const std::vector<int> & v_init)
: v(v_init),
minmax_v(*(std::min_element(v.begin(), v.end())), *(std::max_element(v.begin(), v.end()))) {}
};
It is possible to somehow use std::minmax_element
to initialize minmax_v
? The idea is is to improve performance.
Of course, one could use std::minmax_element
in the constructor body and drop the const
for minmax_v
but that doesn’t feel right either.
Upvotes: 4
Views: 275
Reputation: 9406
Since you are asking this to improve performance, I want to point out that it may be faster to use a custom loop in this case. In your example, there are two operations which both iterate over the vector: the initialization of v
and std::minmax_element
both traverse the input vector. Iterating twice seems not optimal, so it may be better to merge the operations into one loop.
class MyClass {
private:
MyClass(std::tuple<int, int, std::vector<int>>&& x): v( std::get<2>(x) ), minmax_v( std::get<0>(x), std::get<1>(x)) {
}
static std::tuple<int, int, std::vector<int>> init(std::vector<int> const& v_init) {
std::vector<int> v;
v.reserve(v_init.size());
int min = v_init.front();
int max = v_init.front();
for(auto const& x: v_init) {
v.push_back(x);
if (min > x) {
min = x;
}
if (max < x) {
max = x;
}
}
return std::make_tuple(min,max, std::move(v));
}
public:
const std::vector<int> v;
const std::pair<int, int> minmax_v;
MyClass(const std::vector<int> & v_init): MyClass(init(v_init)) {}
};
I am assuming that the input vector is not empty as the original example also dereferences the result std::min_element
and std::max_element
without checking it against v_init.end()
. Since const
members cannot be initialized in the constructor's body I had to come up with a work-around using a helper tuple.
I also want to say that I would probably prefer using STL algorithms instead of the hand-written loop until I can see a performance issue. Modern compilers probably merge the two loop into a single loop and produce efficient code.
Upvotes: 1
Reputation: 217235
Use a delegating constructor to obtain the pair of minmax_element()
results in a single argument:
class MyClass {
public:
const std::vector<int> v;
const std::pair<int, int> minmax_v;
MyClass(const std::vector<int> & v_init)
: MyClass(v_init, std::minmax_element(v_init.begin(), v_init.end()))
{}
private:
MyClass(const std::vector<int> & v_init,
std::pair<std::vector<int>::const_iterator,
std::vector<int>::const_iterator> p)
: v(v_init), minmax_v(*p.first, *p.second)
{}
};
Upvotes: 7