Why does this program using Collections.sort only fail for lists of size 32 or more?

Question

The following program throws the following exception:

java.lang.IllegalArgumentException: Comparison method violates its general contract!

Answer 1

It depends on the implementation, but in openjdk 8 the size of the array is checked against MIN_MERGE, which is equal to 32. This avoids the call to mergeLo/mergeHi which throw the exception.

From JDK / jdk / openjdk / 7u40-b43 8-b132 7-b147 - 8-b132 / java.util.TimSort:

static <T> void sort(T[] a, int lo, int hi, Comparator<? super T> c,
                     T[] work, int workBase, int workLen) {
    assert c != null && a != null && lo >= 0 && lo <= hi && hi <= a.length;

    int nRemaining  = hi - lo;
    if (nRemaining < 2)
        return;  // Arrays of size 0 and 1 are always sorted

    // If array is small, do a "mini-TimSort" with no merges
    if (nRemaining < MIN_MERGE) {
        int initRunLen = countRunAndMakeAscending(a, lo, hi, c);
        binarySort(a, lo, hi, lo + initRunLen, c);
        return;
    }

    /**
     * March over the array once, left to right, finding natural runs,
     * extending short natural runs to minRun elements, and merging runs
     * to maintain stack invariant.
     */
    TimSort<T> ts = new TimSort<>(a, c, work, workBase, workLen);
    int minRun = minRunLength(nRemaining);
    do {
        // Identify next run
        int runLen = countRunAndMakeAscending(a, lo, hi, c);

        // If run is short, extend to min(minRun, nRemaining)
        if (runLen < minRun) {
            int force = nRemaining <= minRun ? nRemaining : minRun;
            binarySort(a, lo, lo + force, lo + runLen, c);
            runLen = force;
        }

        // Push run onto pending-run stack, and maybe merge
        ts.pushRun(lo, runLen);
        ts.mergeCollapse();

        // Advance to find next run
        lo += runLen;
        nRemaining -= runLen;
    } while (nRemaining != 0);

    // Merge all remaining runs to complete sort
    assert lo == hi;
    ts.mergeForceCollapse();
    assert ts.stackSize == 1;
}

Answer 2

Java 8 uses TimSort algorithm to sort the input. In TimSort there is a merging phase that happens when the length is at least 32. When the length is lower than 32 then a simple sorting algorithm is used that probably doesn't detect the contract violation. Let the source code comments of TimSort.java speak for itself:

class TimSort<T> {
    /**
     * This is the minimum sized sequence that will be merged.  Shorter
     * sequences will be lengthened by calling binarySort.  If the entire
     * array is less than this length, no merges will be performed.
     *
     * This constant should be a power of two.  It was 64 in Tim Peter's C
     * implementation, but 32 was empirically determined to work better in
     * this implementation.  In the unlikely event that you set this constant
     * to be a number that's not a power of two, you'll need to change the
     * {@link #minRunLength} computation.
     *
     * If you decrease this constant, you must change the stackLen
     * computation in the TimSort constructor, or you risk an
     * ArrayOutOfBounds exception.  See listsort.txt for a discussion
     * of the minimum stack length required as a function of the length
     * of the array being sorted and the minimum merge sequence length.
     */
    private static final int MIN_MERGE = 32;