The "date" encoding is better than "dic" encoding as it doesn't need storage cost.
Did you try this (explicitly let SQL parser know it is a Date value): where "DATE">=DATE'2016-01-01' 2016-08-09 17:38 GMT+08:00 张天生 <[email protected]>: > I found the answer, it because DATE dimension rowkey encode is date. When > executes "select APPID, SUM(IMPS) as imps from EXT_MID_EVENT_JOIN where > "DATE">=1462060800000 and "DATE"<1462147200000 group by APPID order by imps > desc limit 100", it returned right results. Btw, whether the date rowkey > encode is better performace than dictionary encode? > > 张天生 <[email protected]>于2016年8月9日周二 下午3:55写道： > >> I built a cube, it has 2 dimesions : CAMPAIGNID, DATE, and has 2 >> measures: count(1), topN(100). TOP-N is grouped by APPID and order by/sum >> by IMPS measure. When i queried: select APPID, SUM(IMPS) as imps from >> EXT_MID_EVENT_JOIN where "DATE">='2016-01-01' and "DATE"<'2016-01-08' group >> by APPID order by imps desc limit 100, it returned 0 records. Abviously >> there are a lot of records. Can someone explain why it always returned 0 >> records? >> > -- Best regards, Shaofeng Shi
