Wikipedia:Articles for deletion/Several Unique Sort

Several Unique Sort
The only Google hits I get for "several unique sort" are Wikipedia mirrors, pages from critticall.com and a forum post. This article is nothing but an advertisement for Critticall, and (most significantly) it violates the no original research policy. Fredrik | talk 23:17, 20 Nov 2004 (UTC)
 * Delete per reasons above. -Fennec (&#12399;&#12373;&#12400;&#12367;&#12398;&#12365;&#12388;&#12397;) 04:31, 21 Nov 2004 (UTC)
 * Delete --fvw* 06:24, 2004 Nov 21 (UTC)
 * I just tried stepping through the code, and it doesn't seem to even work. Delete as original research. -- Cyrius|&#9998; 21:39, 21 Nov 2004 (UTC)

It works fine, and it doesn't matter what it _seems_ to you. And it is a separate algorithm from others. I suggest you try, before you talk.


 * I did try, and it didn't work. I said "seem" because I reserved the possibility that my manual code analysis was flawed. -- Cyrius|&#9998; 15:07, 24 Nov 2004 (UTC)


 * I just retried stepping through it. Although the algorithm does indeed avoid excess swapping, the implementation is broken. Try sorting a list of two elements using it, it never terminates the outer do loop because when it tries to sort the single first element, swapindex and position both get set to zero and never leave that value.


 * Even if it's fixed, the algorithm is still fundamentally a bubble sort, publishing this is still original research, and quite frankly, I think the whole article is an advertisement. -- Cyrius|&#9998; 16:00, 24 Nov 2004 (UTC)


 * Well, you tried it wrong, it _is_ working. And the Quick Sort is also "fundamentally a bubble sort".


 * Advertisement? You will have a lot to clean, if that's the only problem.


 * Just checked, it is working. And it is the fastest, when there is not a lot of unique records. By far. (comments by User:212.30.90.228)


 * Comment: Cyrius's analysis is mistaken; the algorithm terminates and sorts arrays of two items. Gazpacho 02:22, 25 Nov 2004 (UTC)
 * Yep, I got fooled by the rather convoluted structure. Once I figured out where my mistake was, I rewrote the algorithm into a more conventional form. -- Cyrius|&#9998; 03:52, 25 Nov 2004 (UTC)
 * Delete: Cyrius's analysis of the algorithm may be mistaken, but I think his analysis of the article is spot-on: it's original research which is only mildly interesting (not notable) because of how this minor improvement to the very non-state-of-the-art bubble sort was discovered.  -- Antaeus Feldspar 02:40, 25 Nov 2004 (UTC)
 * Improvements are so "minor", that this sort is faster then QuickSort, when you have 1 million records and only several of them unique. Again, test it before you judge it. (comment by User:212.30.90.228)


 * And then again, what's the use of opinions of those, who are unable to actually test a piece of code? Who have only some "general remarks"?


 * You say "faster than Quicksort" like it's some sort of revolutionary development. Quicksort is known to have problems with such datasets. -- Cyrius|&#9998; 19:17, 25 Nov 2004 (UTC)


 * Quicksort is NOT known to have problems with such datasets. QuickSort sorts that kind of arrays with several unique records just fine! SeveralUniqueSort is faster in this case, however.


 * The hell it isn't! An array where every element is the same is one of the known triggers for Quicksort's worst case N^2 performance. An array with many identical elements will eventually contain one or more subarrays that are wholly identical, which will again cause Quicksort to throw fits.


 * This assumes an ordinary Quicksort, as the algorithm can be modified to better deal with identical elements. -- Cyrius|&#9998; 22:10, 25 Nov 2004 (UTC)


 * We are NOT talking about the naive version of QuickSort. We are talking about the so called introsort, for example. Where this bad case is eliminated from QuickSort and which behaves very good on the semi sorted arrays. Its weak point is elsewhere, in fact it is very good for these "several unique cases". But the Several Unique is much better for those cases than ANY other sort. Not just QuickSort.


 * Keep: This is indeed original research, which is why I added the warning to the top of the article. However, the algorithm described is hardly someone's pet theory – it actually does what it claims to do and would be of interest to  anyone researching sorting algorithms.  Wikipedia is supposed to cover the sum of human knowledge, and while Several Unique may be new the article is still genuinely informative.  &mdash; Lee J Haywood 10:01, 26 Nov 2004 (UTC)
 * Wikiepdia has an official policy of no original research. Are you saying we should revoke that policy? -- Cyrius|&#9998; 20:53, 26 Nov 2004 (UTC)
 * Of course not, but the policy talks specifically about theories. I don't think that an unambiguous algorithm counts as a theory by itself.  The discussion about its relative performance and usefulness does count as theory, and is speculative while the algorithm is new, but I maintain that the article is fine so long as a warning is present.  &mdash; Lee J Haywood 22:50, 26 Nov 2004 (UTC)
 * While it talks about theories, it is not limited solely to theories. An article about a new algorithm "introduces original ideas". -- Cyrius|&#9998; 23:11, 26 Nov 2004 (UTC)
 * Delete as primary research (see What Wikipedia is not). Niteowlneils 16:53, 26 Nov 2004 (UTC)


 * And unedelete it when? Never? After it is published elsewhere (where)? After you get more comfortable with the (digital) evolution (of code and/or algorithms)? When you will understand how it works? After a patent for it is granted?
 * Yes. We pride ourselves on being hidebound reactionaries, you see; all our rules are based around our comfort level.  We begrudgingly accepted that maybe this new-fangled "fire" had its places sometime last year.  We are of course also rock-stupid; we could not possibly understand how your wonderful, never-before-equalled, dazzling algorithm works because we are still not sure even how the little bunnies who live inside our computers manage to light up all those little pixels at just the right times.  yes, you are completely right; this VfD only exists because we are hidebound and stupid -- and not because Wikipedia has a purpose that does not include being a vehicle for publishing original research. -- Antaeus Feldspar 16:18, 27 Nov 2004 (UTC)
 * If you say so ... But I am glad, that the only reason now is the "primary research". That those "clever analysis" are out.