> In my opinion, kylin can change the yyyy-MM-dd and yyyyMMdd to millis. and the cardinality is equal to the cardinality of 'yyyyMMdd' or yyyy-MM-dd.
Makes sense. The data format is a separate thing as to cardinality. By rounding millis to hours to days, the column can still be low in cardinality. I'm re-opening the JIRA 1698. On Thu, May 19, 2016 at 5:03 PM, 耳东 <775620...@qq.com> wrote: > we have opened a JIRA. And the reponse is as follows: > millis format as partition column will not be support in short run. The > reason is that millis format tends to come with large cardinality, which is > not acceptable by kylin. As a workaround you may use hive view to map the > millis format to 'yyyy-MM-dd' or 'yyyyMMdd' format. > In my opinion, kylin can change the yyyy-MM-dd and yyyyMMdd to millis. > and the cardinality is equal to the cardinality of 'yyyyMMdd' or > yyyy-MM-dd. > > > > > > > > ------------------ 原始邮件 ------------------ > 发件人: "Li Yang";<liy...@apache.org>; > 发送时间: 2016年5月19日(星期四) 下午4:49 > 收件人: "dev"<dev@kylin.apache.org>; > > 主题: Re: Partition date column > > > > Please open a JIRA for the new feature. It should not be too difficult to > support long millis as partition date column. > > On Tue, May 17, 2016 at 12:25 AM, 耳东 <775620...@qq.com> wrote: > > > Hi all : > > When we create the model, there is a property called partition date > > column, and the format the web interface shows is yyyyMMdd or yyyy-MM-dd. > > Can this partition date column be a timestamp like 1463406381 or > > 1463415185143. > > > > > > When we build the cube, there are two parameters called startTime > and > > endTime ,and the format the instruction shows is 1388563200000 for > > 2014-1-1. Which column will be used to filter the data? If it use the > > partition date column, is there any tranformation in the kylin? >
