[Mondrian] are readTuple queries really necessary ?

[Julian Hyde]

As Luc suggests, we want to do better. If we determine that we need a tuple query (because of a NONEMPTY axis, for instance) it would be nice if a single SQL query retrieved both the cells and the members. (This takes some pretty serious query-optimization; we’re working on it.)

On Jan 30, 2014, at 6:41 AM, Luc Boudreau <lucboudreau at gmail.com> wrote:

> Hello Arvind,
> 
> We've been discussing merging the segment and tuple queries for some time now, but their separation is at the core of the algorithm which resolves a query into its result. It isn't a trivial matter to merge the two together, and so far it hasn't happened.
> 
> As for the tuple queries using the fact table, this usually happens when an MDX statement asks for the non empty tuples. In that situation, we join through the fact table and voila. If the crossjoin is simple enough, the fact table should be omitted from the query.
> 
> Luc
> 
> 
> On Thu, Jan 30, 2014 at 2:42 AM, Arvindkumar J <j.arvindkumar at gmail.com> wrote:
> Hi All, 
> 
> I have a query regarding SQL query generation. 
> 
> For each MDX query, mondrian fires set of readTuple queries and segment.load queries. 
> I understand that readTuple queries are used to populate the axes and segment load queries to populate the cell values. So usually there are more readTuple queries compared to segment.load queries.
> 
> Why can't we avoid the readTuple queries in case where the predicates are fixed ? 
> (i know we can't avoid in-case the predicates are like [Product].children ). 
> Instead, we can populate the axes from the input predicates and filtering of tuples with 0 values can be doneter the segment load queries. 
> 
> The reason i'm asking this is the readTuple queries are taking major part of the query time. And i notice these queries are on fact tables. Are these queries supposed to be on fact tables or only dimension tables ?
> 
> Please correct me if my understanding is not correct. 
> 
> Thanks
> Arvind
> 
> 
> _______________________________________________
> Mondrian mailing list
> Mondrian at pentaho.org
> http://lists.pentaho.org/mailman/listinfo/mondrian
> 
> 
> _______________________________________________
> Mondrian mailing list
> Mondrian at pentaho.org
> http://lists.pentaho.org/mailman/listinfo/mondrian

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.pentaho.org/pipermail/mondrian/attachments/20140130/122560ec/attachment.html