I don't see my other arguments being convincingly refuted, so they still hold.
How ancient is your machine? Quicksorting (2^31 - 16) elements (because that's Java's hardcoded limit) takes < 11 seconds on my machine, and a big chunk of that time is taken in the random number generation to create the input...
# Temp.java
import java.util.*;
public class Temp { public static void main(String[] args) { byte[] arr = new byte[0x7FFFFFF0]; new Random().nextBytes(arr); Arrays.sort(arr); } }
$ javac Temp.java && time -p java Temp
real 10.30What article are you referring to? This article is about binary search and mergesort, not quicksort?
And which quicksort has O(n^2) typical behavior? That's the worst-case behavior you normally only get on adversarial inputs, not the typical one. (And standard libraries have workaround for that anyway.)
I still think that it's normal for programs to behave weird when the input parameters are getting close to INT_MAX. Sometimes it's unavoidable. And if it's not specified in the function docs, you should go and check the implementation as a programmer. For binary search it is avoidable, so the criticism of the linked implementation is fair.