I was thinking more of general (not just LIMIT 1) optimization: http://en.wikipedia.org/wiki/Partial_sorting
On Sun, Apr 7, 2013 at 6:16 PM, Simon Slavin <slav...@bigfraud.org> wrote: > > On 7 Apr 2013, at 3:51pm, Stephen Chrzanowski <pontia...@gmail.com> wrote: > > > I don't know if it'd be an interesting optimization. Who's to say what > the > > order ends up as prior to the sort? Take for example if I have a list of > > dollars and cents being returned from a query. I want the 5 top highest > > cost items out of 6 possibilities, if I keep the top 5 unsorted, item 6 > > could have been the very top of the list. Thereby making the desired > > result wrong. > > The optimization I was thinking of was simpler than that. An example > would have been ORDER BY score DESC LIMIT 1. Theoretically the engine > could notice that all it needed was the top-scoring result, and just do a > quick scan for the highest score, which would be less effort than sorting > all the rows into order. I can imagine one of the big heavy-duty SQL > engines doing this but perhaps not SQLite because it would require a lot of > extra code and introduce a lot of test cases. > > A server/client SQL system can do a kind of progressive indexing because > it runs on a server and can do a lot of processing without tying up the > client computer. So with the above type of SELECT the engine will return > the first result as the client asks for it. Then it will immediately try > to find the second result, then the third, etc. aiming to have each result > ready before the client asks for it. I believe MySQL does this under > certain conditions. > > Simon. > _______________________________________________ > sqlite-users mailing list > sqlite-users@sqlite.org > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > -- ˙uʍop-ǝpısdn sı ɹoʇıuoɯ ɹnoʎ 'sıɥʇ pɐǝɹ uɐɔ noʎ ɟı _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
