One correction: from the code I see, when the extended column is not determined, it will use the first appeared value; In your case, the first value might be all blank, so it appeared like all empty. The related code is: https://github.com/apache/kylin/blob/master/core-metadata/src/main/java/org/apache/kylin/measure/extendedcolumn/ExtendedColumnMeasureType.java#L281
This measure has such limitations, please not use it if you are not sure about it. I'm thinking whether we need to disable it. ShaoFeng Shi <shaofeng...@apache.org> 于2018年11月7日周三 上午9:04写道： > Hi Leo, > > "Extended column" is a special measure, which is to store a dimension > value as a measure. It doesn't support aggregation, so if there are > multiple values in this "measure", it will be empty. > > Let me give an example. > > Say you have a table, which has dimensions like "*user_id*", "*user_name*", > "*age*"; Here the "*user_id*" is unique, which will be used to do > filtering/searching; The "user_name" will not be used for filtering, but it > is nice to show the value in the result. So, in this case, we can declare > "user_id" and "age" as dimensions, but define "user_name" as an extended > column. Then the cube will have 2 dimensions instead of 3, its size will be > smaller. > > When you run a query like "select user_id, user_name, count(*) from > fact_table group by user_id, user_name", it will be expected, because the > "user_name" value is decided (明确的) under the context of a "user_id"; > > But if you run a query without "user_id", for example, "select age, > user_name, count(*) from fact_table group by age, user_name", this column > will be empty, because it has no "user_id" in the row. The extended column > can not work in the query independently > > So, the conclusion is, the extended column(s) need to appear together with > its "host" column. If its "host" column is absent, the value will be empty. > > We will add this to Kylin document. > > > > Francis Liang <soulvagr...@hotmail.com> 于2018年11月5日周一 下午5:21写道： > >> Actually you need at least one aggregation, such as sum(lost_day_cnt), or >> you may try select a, b from table group by a, b, which should not be the >> normal case however. Hope this makes sense. Best, Feng. >> >> >> >> *发件人: *宋海涛 <hts...@gaojihealth.com> >> *发送时间: *2018年11月5日 17:16 >> *收件人: *user@kylin.apache.org >> *主题: *答复: Extend Column Is Empty >> >> >> >> select zone, data_from,ent_name,org_no, org_name, goods_id,goods_spec, >> big_mc,mid_mc,sma_mc,mini_mc,lost_day_cnt, >> >> lost_amt,avg_paid_in_amt,dt >> >> from gjst.st_goods_lost_monitor_detail_day where dt='20181102' and >> >> zone='西南平台' and data_from='haitang' and org_no='16208' >> >> >> >> *发件人:* Francis Liang [mailto:soulvagr...@hotmail.com] >> *发送时间:* 2018年11月5日 17:14 >> *收件人:* user@kylin.apache.org >> *主题:* 答复: Extend Column Is Empty >> >> >> >> Could you share your sql statement? Thanks. Best, Feng. >> >> >> >> *发件人**:* 宋海涛 <hts...@gaojihealth.com> >> *发送时间**:* Monday, November 5, 2018 4:58:53 PM >> *收件人**:* user@kylin.apache.org >> *主题**:* Extend Column Is Empty >> >> >> >> Hi guys, >> >> >> >> When I run a SQL which contains extended columns, I got empty >> values for extended columns . Anyone knows why? >> >> For example , ENT_NAME and ORG_NAME are extended column based on >> DATA_FROM and ORG_NO: >> >> [image: cid:image001.png@01D4752B.489B5550] >> >> >> >> Thanks a lot. >> >> >> >> Leo >> >> >> > > > -- > Best regards, > > Shaofeng Shi 史少锋 > > -- Best regards, Shaofeng Shi 史少锋
