Hi, Top-N is usefull for one 'Top 10', but can be useful know the Sum of 'the Others' (= sum (top > 10))
Example: In a Shop Top 10 Sellers, sold 1.2M $ How much sold 'the others'? 1.2M $ its a lot respect the others? I know that this is not easy to implement, but if someboy have any idea ... 2016-08-09 6:19 GMT+02:00 ShaoFeng Shi <[email protected]>: > Hi Tiansheng, > > The less post-aggregation, the better query performance; So for a specific > query, if the "signle groupby column topN" need further aggregation to get > the final result, but "multiple groupby column topN" doesn't, then the > later one would have better performance. > > I didn't compare that, just personal cents; Welcome to do benchmark and > share with the community :-) > > 2016-08-09 11:54 GMT+08:00 张天生 <[email protected]>: > >> I have a question: whether multiple column groupby is better performance >> than a single column groupby in topN measure. As i known it all can agg >> other dimensions. >> Whether it there was performance optimization in mulitple column groupby >> in topN measure. >> >> ShaoFeng Shi <[email protected]>于2016年8月8日周一 下午6:20写道： >> >>> Alberto is correct; SUM(1) and multiple columns are implemented in Kylin >>> core, but from UI you couldn't define that; You need manually edit metadata >>> for that. >>> >>> 2016-08-08 18:02 GMT+08:00 赵天烁 <[email protected]>: >>> >>>> ok,I'll have a try >>>> >>>> ------------------------------ >>>> >>>> 赵天烁 >>>> >>>> Kevin Zhao >>>> >>>> *[email protected] <[email protected]>* >>>> >>>> >>>> >>>> 珠海市魅族科技有限公司 >>>> >>>> MEIZU Technology Co., Ltd. >>>> >>>> 广东省珠海市科技创新海岸魅族科技楼 >>>> >>>> MEIZU Tech Bldg., Technology & Innovation Coast >>>> >>>> Zhuhai, 519085, Guangdong, China >>>> >>>> meizu.com >>>> >>>> >>>> *From:* Alberto Ramón <[email protected]> >>>> *Date:* 2016-08-08 17:59 >>>> *To:* [email protected] >>>> *CC:* ShaoFeng Shi <[email protected]> >>>> *Subject:* Re: Re: does kylin support top-N on a count or count >>>> distinct measure? >>>> In teorical en v1.5.3, you can Group by 'n' columns: >>>> https://issues.apache.org/jira/browse/KYLIN-1693 >>>> >>>> I don't tested 1.5.3 yet, and I don't know if has been implemented in >>>> UI Kylin, perhaps you can add this columns to JSON manually :) >>>> >>>> BR, Alberto >>>> >>>> 2016-08-08 11:37 GMT+02:00 赵天烁 <[email protected]>: >>>> >>>>> SUM(1)? you mean just left ORDER|SUM by Column empty? ,then another >>>>> prob is I can't configure more than one group by column to it,how to walk >>>>> around that? >>>>> >>>>> ------------------------------ >>>>> >>>>> 赵天烁 >>>>> >>>>> Kevin Zhao >>>>> >>>>> *[email protected] <[email protected]>* >>>>> >>>>> >>>>> >>>>> 珠海市魅族科技有限公司 >>>>> >>>>> MEIZU Technology Co., Ltd. >>>>> >>>>> 广东省珠海市科技创新海岸魅族科技楼 >>>>> >>>>> MEIZU Tech Bldg., Technology & Innovation Coast >>>>> >>>>> Zhuhai, 519085, Guangdong, China >>>>> >>>>> meizu.com >>>>> >>>>> >>>>> *From:* ShaoFeng Shi <[email protected]> >>>>> *Date:* 2016-08-08 11:32 >>>>> *To:* user <[email protected]> >>>>> *Subject:* Re: does kylin support top-N on a count or count distinct >>>>> measure? >>>>> For sorting on count, you can use SUM(1) as the expression; >>>>> >>>>> For sorting on other measure, it is on roadmap: https://issues.apache >>>>> .org/jira/browse/KYLIN-1377 >>>>> >>>>> We welcome the community to contribute on such enhancements, anyone >>>>> want to have a try? >>>>> >>>>> 2016-08-05 15:24 GMT+08:00 赵天烁 <[email protected]>: >>>>> >>>>>> right now top-N measure need to specify a sum column, >>>>>> does kylin support top-N on a count or count distinct measure? >>>>>> >>>>>> ------------------------------ >>>>>> >>>>>> 赵天烁 >>>>>> >>>>>> Kevin Zhao >>>>>> >>>>>> *[email protected] <[email protected]>* >>>>>> >>>>>> >>>>>> >>>>>> 珠海市魅族科技有限公司 >>>>>> >>>>>> MEIZU Technology Co., Ltd. >>>>>> >>>>>> 广东省珠海市科技创新海岸魅族科技楼 >>>>>> >>>>>> MEIZU Tech Bldg., Technology & Innovation Coast >>>>>> >>>>>> Zhuhai, 519085, Guangdong, China >>>>>> >>>>>> meizu.com >>>>>> >>>>> >>>>> >>>>> >>>>> -- >>>>> Best regards, >>>>> >>>>> Shaofeng Shi >>>>> >>>>> >>>> >>> >>> >>> -- >>> Best regards, >>> >>> Shaofeng Shi >>> >>> > > > -- > Best regards, > > Shaofeng Shi > >
