Split that into two cubes: in one it acts as a dimension, in the other, it acts as count distinct measure.
2017-09-29 11:12 GMT+08:00 崔苗 <cuim...@danale.com>: > if a column was used both as dimension and distinct count,what should be > use to encode ? > 在 2017-09-29 11:05:28，ShaoFeng Shi <shaofeng...@apache.org> 写道： > >Use GD for dimension encoding will cause query inaccurate, that should be > >avoided. > > > >Please use dict or fixed_length encoding for that column. > > > >2017-09-29 10:37 GMT+08:00 柯南 <zzm_b...@qq.com>: > > > >> Yes，I am careful to use GD. Could I have another way to instead of GD？ > >> > >> > >> > >> > >> ------------------ 原始邮件 ------------------ > >> 发件人: "ShaoFeng Shi";<shaofeng...@apache.org>; > >> 发送时间: 2017年9月29日(星期五) 上午9:41 > >> 收件人: "dev"<dev@kylin.apache.org>; > >> > >> 主题: Re: A problem about Global Dictionary and cube merge > >> > >> > >> > >> I see; Did you use GlobalDictionaryBuilder for a dimension column? The > GD > >> should only be used for count distinct measure, not suitable for > dimension > >> encoding. > >> > >> 2017-09-28 23:57 GMT+08:00 柯南 <zzm_b...@qq.com>: > >> > >> > Thank you for your reply! > >> > I know this command.But Because its segment could not be merged, > even if > >> > run this command ,the hbase tables could not be cleaned. > >> > > >> > > >> > > >> > > >> > ------------------ 原始邮件 ------------------ > >> > 发件人: "ShaoFeng Shi";<shaofeng...@apache.org>; > >> > 发送时间: 2017年9月28日(星期四) 晚上9:29 > >> > 收件人: "dev"<dev@kylin.apache.org>; > >> > > >> > 主题: Re: A problem about Global Dictionary and cube merge > >> > > >> > > >> > > >> > You need run this: > >> > https://kylin.apache.org/docs21/howto/howto_cleanup_storage.html > >> > > >> > 2017-09-28 20:40 GMT+08:00 柯南 <zzm_b...@qq.com>: > >> > > >> > > hi,all： > >> > > Because of our yarn and hbase`s version limited,we deploy > apache > >> > > kylin 1.6.0. For the correct result of 'count(distinct A_col)' > across > >> > > segments, we must use Global Dictionary to A_col. When the cube > contain > >> > > configuration of 'org.apache.kylin.dict.GlobalDictionaryBuilder', > its > >> > > segment could not be merged.(AppendTrieDictionary class don`t > support > >> > > getValueBytesFromIdImpl method and error is below) We will build the > >> cube > >> > > every day, that means a new hbase table will be created every day. > And > >> > the > >> > > segment could not be merged, that means the hbase tables could not > be > >> > > cleaned. That`t terrible.Give me some advice? > >> > > Thank you! Looking forward to your reply. > >> > > >> > > >> > > >> > > >> > -- > >> > Best regards, > >> > > >> > Shaofeng Shi 史少锋 > >> > > >> > >> > >> > >> -- > >> Best regards, > >> > >> Shaofeng Shi 史少锋 > >> > > > > > > > >-- > >Best regards, > > > >Shaofeng Shi 史少锋 > > > -- Best regards, Shaofeng Shi 史少锋
