[
https://issues.apache.org/jira/browse/KYLIN-1788?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15385604#comment-15385604
]
Wang Cheng commented on KYLIN-1788:
-----------------------------------
Essentially it is not a bug but a strategy problem. Of course it can be solved
by tuning the threshold "maxSize", however, it's difficult to get an
appropriate size.
A compromise proposal is to count "mandatoryDims.size()" as 1 if its size isn't
zero:
{code:java}
int mandatorySize = mandatoryDims.size() > 0 ? 1 : 0 ;
if ( mandatorySize + normalDimSize + hierarchySize + jointSize > maxSize) {
context.addResult(ResultLevel.ERROR, "Aggregation group " + index + "
has too many dimensions");
continue;
}
{code}
> Allow arbitrary number of mandatory dimensions in one aggregation group
> ------------------------------------------------------------------------
>
> Key: KYLIN-1788
> URL: https://issues.apache.org/jira/browse/KYLIN-1788
> Project: Kylin
> Issue Type: Bug
> Affects Versions: v1.5.2
> Reporter: hongbin ma
> Assignee: hongbin ma
>
> To prevent one aggregation group containing too many combinations we apply a
> check
> {code:java}
> if (mandatoryDims.size() + normalDimSize + hierarchySize + jointSize >
> maxSize) {
> context.addResult(ResultLevel.ERROR, "Aggregation group " +
> index + " has too many dimensions");
> continue;
> }
> {code}
> however the formular fails to take into account the case where there're many
> mandatory dimensions. For example, if we have 50 dimensions in a cube and we
> only need the base cuboid, then what we want is a single aggregation group
> containing all the dimensions, each of them being a mandatory.
> since mandatory dimensions are more "encouraged", I suggest to remove
> counting mandatory dimensions in the formula. the revised code will be:
> {code:java}
> if (normalDimSize + hierarchySize + jointSize > maxSize) {
> context.addResult(ResultLevel.ERROR, "Aggregation group " +
> index + " has too many dimensions");
> continue;
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
