Slicing a C-array creating a C++ vector, keeping only 2 elements every 3

Question

I have a C-array like this:

int X[]={0, 1, 2, 3, 4, 5, 6, 7, 8};

I need to create two stl vectors out of this array, by means of slicing, possibly sharing the highest amount of memory and making less deep copies possible. The first vector Y must contain only the 0-th and first element each three elements, for example in this case this new vector Y will contain

std::vector<int> Y; // contains: [0,1,3,4,6,7]

and another vector Z must contain each 3rd element in the original array:

std::vector<int> Z; // contains [2,5,8]

A first solution based on for loops with copy is the following:

vector<int> Y,Z;
for (int i=0; i<9;i+=3)
{
    Y.push_back(*(X+i));
    Y.push_back(*(X+i+1));
    Z.push_back(*(X+2));
}

but I'm pretty sure that by means of custom iterators the problem could have a more efficient solution. Is there, anyway, some other faster version of implementing a mask-view on this array that avoid generating copies?

I don't think there's a way to do this with custom iterators without heavy overengineering. — krzaq, Oct 04 '16 at 13:48
I don't see how custom iterators could make your simple for loop more "efficient" (although you could do with some `.reserve`s) https://en.wikipedia.org/wiki/KISS_principle — Lightness Races in Orbit, Oct 04 '16 at 13:51
Other than using `reserve()` I doubt you could get much more efficient than that. — Galik, Oct 04 '16 at 13:52
@LightnessRacesinOrbit: Isn't "Simplify, then add lightness" a better fit for you :-) — davidhigh, Oct 04 '16 at 14:27
Efficient how? Relative to what? In creation or in use of Y and Z? Do you need future changes to X to be reflected in Y and Z? How big is X? — Peter, Oct 04 '16 at 15:10

score 1 · Answer 1 · answered Oct 04 '16 at 14:36

Do you maybe mean something like this:

#include <iostream>
#include <vector>
using namespace std;

struct MyIndexing {
    int* original;
    MyIndexing(int* o) : original(o) {}
    int& getY(int index){ return original[0]; /* to be done... */ }
    int& getZ(int index){ return original[(index+1)*3 -1]; }
};

int main() {
    int X[]={0, 1, 2, 3, 4, 5, 6, 7, 8};
    MyIndexing mi(X);
    for (int i=0;i<3;i++){ std::cout << mi.getZ(i) << " "; }
    return 0;
}

Prints:

2  5  8

It does not create vectors, but anyhow your requirements are a bit contradicting. Either you avoid copies or you create new vectors. Both isnt possible.

score 0 · Answer 2 · answered Oct 04 '16 at 14:26

0

If you HAVE to create vectors, you MUST make a copy of the data, as the vectors data is contiguous.

If you don't need the vectors, you could create 2 interfaces (forgive my C# terminology, but I think it's applicable), one that would provide a getter for Y data, one that would offer for X data; the implementation, extending these 2 interfaces can then hide the fact that your data is in a vector at all.

answered Oct 04 '16 at 14:26

UKMonkey

6,941
3
21
30

All factually true. However, such an interface would be quite inefficient to use in most scenarios I'm imagining. Still, that does appear to be what the OP is (now) asking for, post-edit. You didn't really provide a solution though – Lightness Races in Orbit Oct 04 '16 at 14:30

Slicing a C-array creating a C++ vector, keeping only 2 elements every 3

2 Answers2