Weighted probability with long doubles

Question

I am working with an array of roughly 2000 elements in C++.

Each element represents the probability of that element being selected randomly.

I then have convert this array into a cumulative array, with the intention of using this to work out which element to choose when a dice is rolled.

Example array: {1,2,3,4,5}

Example cumulative array: {1,3,6,10,15}

I want to be able to select 3 in the cumulative array when numbers 3, 4 or 5 are rolled.

The added complexity is that my array is made up of long doubles. Here's an example of a few consecutive elements:

0.96930161525189592646367317541056252139242133125662803649902343750 0.96941377254127855667142910078837303444743156433105468750000000000 0.96944321382974149711383993199831365927821025252342224121093750000 0.96946143938926617454089618153290075497352518141269683837890625000 0.96950069444055009509463721739663810694764833897352218627929687500 0.96951751803395748961766908990966840065084397792816162109375000000

This could be a terrible way of doing weighted probabilities with this data set, so I'm open to any suggestions of better ways of working this out.

Answer 1

You can use partial_sum:

unsigned int SIZE = 5;
int array[SIZE] = {1,2,3,4,5};
int partials[SIZE] = {0};

partial_sum(array, array+SIZE, partials);
// partials is now {1,3,6,10,15}

The value you want from the array is available from the partial sums:

12 == array[2] + array[3] + array[4];

12 == partials[4] - partials[1];

The total is obviously the last value in the partial sums:

15 == partial[4];