Matlab type arrays in C++

Question

How to define arrays in C++/ Open CV as I do in matlab ?

for example:

x=a:b:c;

or

y=linspace(a,b,n);

Good answer for a similar question [here](https://stackoverflow.com/a/27030598/3079302). — iled, Aug 22 '17 at 05:16

mattgately · Accepted Answer · 2013-08-06T15:26:13.827

9

Refer the previous answers for general answers to your question.

Specifically, to address the two examples that you mention, here is some equivalent c++ code using vectors for dynamically generating the arrays you mentioned (haven't tested):

#include <vector>
using std::vector;

vector<double> generateRange(double a, double b, double c) {
    vector<double> array;
    while(a <= c) {
        array.push_back(a);
        a += b;         // could recode to better handle rounding errors
    }
    return array;
}

vector<double> linspace(double a, double b, int n) {
    vector<double> array;
    double step = (b-a) / (n-1);

    while(a <= b) {
        array.push_back(a);
        a += step;           // could recode to better handle rounding errors
    }
    return array;
}

edited Aug 06 '13 at 15:26

answered Jul 31 '12 at 18:54

mattgately

932
8
22

1

preallocation fills `array` with `n` values, then `push_back` appends about `n` values to the end, so you should drop the preallocation – YuppieNetworking Jul 22 '13 at 15:20
Thanks @YuppieNetworking. I was confusing languages, thinking the constructor allocated without filling. I removed the preallocation, since using `push_back`. – mattgately Aug 06 '13 at 15:28
@mattgately, you can use `reserve` function to be more effective. – Andrey Rubshtein Jul 06 '15 at 08:42

score 3 · Answer 2 · answered Jul 31 '12 at 06:32

OpenCV offers some functions that are similar to Matlab, but their number is very limited.

You can

cv::Mat a = cv::Mat::eye(5);
cv::Mat b = cv::Mat::zeros(5);
cv::Mat img = cv::imread("myGorgeousPic.jpg");
cv::imwrite(img, "aCopyOfMyGorgeousPic.jpg");

It also supports diag()

But for most of that tricky Matlab functionality like linspace or magic or whatever, there is no correspondent in OpenCV, mostly because OpenCV is not a mathematics package, but a computer vision one. If you need some specific function, you can clone it in your project (aka write it by yourself)

score 3 · Answer 3 · edited Apr 13 '17 at 12:53

Unfortunately C++ has nothing inbuilt into it to allow this kind of matrix initialisation. It supports multidimensional arrays but you would need to initialise every element yourself. C++ is a lower level language than Matlab and it involves a lot more work to write the functionality to create and initialise a matrix type variable.

Having said that there are a number of libraries available for use with C++ that make numerical computation easier than if you were to attempt writing it all yourself. If you need to consider using libraries, have a look at this link that suggests a few suitable ones https://scicomp.stackexchange.com/questions/351/recommendations-for-a-usable-fast-c-matrix-library

score 1 · Answer 4 · answered Nov 25 '13 at 15:12

This is a simple implementation if you use openCV.

Mat linspace(double &startP,double &Endp,int &interval)
{
    double spacing = interval-1;
    Mat y(spacing,1,CV_64FC1);
    for (int i = 0; i < y.rows; ++i)
    {
        y.at<double>(i) = startP + i*(Endp - startP)/spacing;
    }
    return y;
}

score 1 · Answer 5 · edited Dec 11 '14 at 00:45

1

double* linspace(int xi, int xf, unsigned int n){
double *vec ;
vec = new double[n];
float esp, falt=xf-xi;
esp=falt/(n-1);
vec[0]=xi;
for(int i=1; i<n; i++)
    vec[i]=vec[i-1]+esp;
return vec;

edited Dec 11 '14 at 00:45

rnrneverdies

15,243
9
65
95

answered Dec 11 '14 at 00:31

Amauro Sand

11
1

1

Your answer should contain an explanation of your code and a description how it solves the problem. – AbcAeffchen Dec 11 '14 at 00:49

Gilad · Answer 6 · 2015-06-14T23:12:37.990

here is my tested linspace very similar to all the others but handles all the cases:

vector<double> Utilities::Linspace(double a, double b, int n) {
    vector<double> array;
    double epsilon = 0.0001;
    double step = (b-a) / (n-1);
    if (a==b)
    {
        for (int i = 0; i < n; i++)
        {
            array.push_back(a);
        }
    }
    else if (step >= 0)
    {
        while(a <= b + epsilon)
        {
            array.push_back(a);
            a += step;           
        }       
    }
    else
    {
        while(a + epsilon >= b )
        {
            array.push_back(a);
            a += step;           
        }       
    }
    return array;
}

Classic error of division by zero when `n=1`...I posted some slight changes (see below) — champost, Jan 04 '16 at 20:34

Poiex · Answer 7 · 2016-05-26T09:08:49.197

1

I would like to add a slight modification to the code proposed by mattgately. I used it and there are some cases when step is not correctly calculated due to division approximation

double step = (b-a) / (n-1);

I just added a small number to the while condition:

while(a <= b+0.00001)

Like this it worked with me and the correct number of intervals was created.

edited May 26 '16 at 09:08

answered May 26 '16 at 08:43

Poiex

138
10

champost · Answer 8 · 2016-04-24T11:18:42.957

0

Extension of @Gilad's answer above for cases when n=0 and n=1 because the latter gives rise to division by zero.

vector<double> linspace(double a, double b, int n) {
    vector<double> array;
    if ((n == 0) || (n == 1) || (a == b))
        array.push_back(b);
    else if (n > 1) {
        double step = (b - a) / (n - 1);
        int count = 0;
        while(count < n) {
            array.push_back(a + count*step);
            ++count;
        }
    }
    return array;
}

edited Apr 24 '16 at 11:18

answered Jan 04 '16 at 20:33

champost

463
4
9

your code doesn't handle the case when step is -ve. (b>a or b – ankittie Jul 11 '17 at 17:40
@ankittie what makes you say that, have you tried the code ? `step` is of `double` type and can be +ve or -ve. – champost Jul 12 '17 at 18:14

Matlab type arrays in C++

8 Answers8