Find median of an array of uints

🧐 Motivation

We are building a reliable price oracle for a decentralized exchange (the DutchX). We would like to take the last n prices, and output the median price. So we need an algorithm that is able to find the median of an array of, say, uints.

📝 Details

A function that takes an array and returns its median. Signature:

function findMedian(uint[] memory array) public Pure Returns (uint)

[Amxx]

Digging up this old topic just to propose an implementation:

function swap(int256[] memory array, uint256 i, uint256 j) internal pure {
    (array[i], array[j]) = (array[j], array[i]);
}
function sort(int256[] memory array, uint256 begin, uint256 end) internal pure {
    if (begin < end) {
    	uint256 j = begin;
    	int256 pivot = array[j];
    	for (uint256 i = begin + 1; i < end; ++i) {
    	    if (array[i] < pivot) {
    	        swap(array, i, ++j);
    	    }
    	}
    	swap(array, begin, j);
    	sort(array, begin, j);
    	sort(array, j + 1, end);
    }
}
function median(int256[] memory array, uint256 length) internal pure returns(int256) {
	sort(array, 0, length);
	return length % 2 == 0 ? Math.average(array[length/2-1], array[length/2]) : array[length/2];
}

[hacker-DOM]

Ended up implementing this by myself 😃 To avoid moving around large arrays in evm, I used a linked list. The code is here https://github.com/gnosis/dx-price-oracle/blob/fcf2dd0fac65754ed1f99f899cb95799892f5d5a/contracts/DutchXPriceOracle.sol#L80

To make it reusable, all that has to be done is:

• Replace the message calls by loading the next array element
• Collapse the use of two arrays of data into one (we use two, one for numerators and one for denominators, because our data are fractions)

Just my two cents ✌️