@shaofeng any insights? On Wed, Aug 3, 2016 at 9:56 AM, Joel Victor <[email protected]> wrote:
> ld think about making this feature optional (A per cube config until we > find a better solution). This feature is useful when data sets are really > large, the size of the distinct count step contributes much more to the > build latency and also where the count step is not complex. > > To add to that the size of the table c > -- Regards, *Bin Mahone | 马洪宾*
