Java 9 and IntSummaryStatistics et al.

Peter Levart peter.levart at gmail.com
Thu Mar 30 14:01:36 UTC 2017



On 03/30/2017 03:14 PM, Chris Dennis wrote:
> This is indeed nice… but I presume that we all agree that the best solution here would be to allow instantiation of an IntSummaryStatistics object in a specific state.

Of course. I just couldn't resist the challenge of Rémi's nice math 
exercise...

Regards, Peter

>
>> On Mar 29, 2017, at 2:43 PM, Peter Levart <peter.levart at gmail.com> wrote:
>>
>>
>>
>> On 03/29/2017 06:06 PM, Chris Dennis wrote:
>>> Remi: I really have to squint pretty hard to see that as anything other than “brute-force” - it’s still an O(N) calculation.
>> Here's O(log2(N)):
>>
>>     static IntSummaryStatistics create(long count, long sum, int min, int max) {
>>         if (count < 2 || min == max) {
>>             return combineProduct(new IntSummaryStatistics(), count, min);
>>         } else {
>>             long minCnt = (count * max - sum) / (max - min);
>>             long maxCnt = count - minCnt - 1;
>>             long minSum = minCnt * min;
>>             long maxSum = maxCnt * max;
>>             long minMaxSum = minSum + maxSum;
>>             int mid = (int) (sum - minMaxSum);
>>             IntSummaryStatistics stats = new IntSummaryStatistics();
>>             combineProduct(stats, minCnt, min);
>>             combineProduct(stats, maxCnt, max);
>>             stats.accept(mid);
>>             return stats;
>>         }
>>     }
>>
>>     static IntSummaryStatistics combineProduct(IntSummaryStatistics stats, long count, int value) {
>>         IntSummaryStatistics pow2 = null;
>>         while (count > 0) {
>>             if (pow2 == null) {
>>                 pow2 = new IntSummaryStatistics();
>>                 pow2.accept(value);
>>             } else {
>>                 pow2.combine(pow2);
>>             }
>>             if ((count & 1L) == 1L) {
>>                 stats.combine(pow2);
>>             }
>>             count >>>= 1;
>>         }
>>         return stats;
>>     }
>>
>>
>> Regards, Peter
>>



More information about the core-libs-dev mailing list