Repeted values on filter (On Analyzer)
I have a hierarchy as follows:
level 1 - Nome (name of person);
Level 2 - Tipo (can be legal or physical).
When I drag the level 2 (type) all kinds of table is presented. My expectation is that it was presented only the "physical" and "legal" alternatives.
In Saiku the filter has the option: "Show unique values", which is exactly what I need. I believe Saiku performs an operation of 'group by' on the bench in this column.
But the analyzer does not have this option.
Any tips on how to resolve this issue in the analyzer?
Maybe [Tipo] should be modeled as a separate dimension, such that you can filter by the two unique values. That way there wouldn't be a separate [Legal] or [Physical] member under each [Nome].
If you'd also like to continue to associate a [Tipo] directly in the [Pessoa] dimension you could also define it as a property of Nome.
Let me try to be clearer, let's use the example of the cube StellWheelsSales that comes in Pentaho 5.0.5:
In Time dimension have three levels
If I drag the Month level to the filter, duplicate values will be displayed in the list as
Now if I drag the first level Years, choose the year 2014, the values of the filter will be presented properly, without being repeated.
For Saiku there is a 'show unique values' option, where if I drag only the Months level values are not presented repeated.
I wonder if anything I can do the mapping, or analyzer, or elsewhere to have the behavior that has Saiku, only exists in the analyzer.
Thanks for help!
Those repeated values are actually each separate, unique dimension members. E.g. [Time]..[Jan], [Time]..[Jan]. There may be a feature in Saiku that allows treating each of the [Jan] members as a single thing, but ultimately it still needs to handle the multiple members (I actually don't see that in the 2.6 version of Saiku-- "Show unique name" just shows the full MDX Unique name for each individual [Jan] member).
My suggestion had been to use a separate [Tipo] dimension such that there truly are just 2 members ([Legal], [Physical]), rather than a unique child Tipo member per [Nome]. From a dimensional modeling perspective it seems odd to me that [Legal] or [Physical] would be *children* of [Nome]. But I don't know your data.
Modeled as a separate dimension? I'm sorry but you completely lost me.How redundant is that from a data store perspective? From an analytical perspective, it should only show the unique values and in this case "physical" and "legal".
Originally Posted by mcampbell
No impact on data store--Not talking about dimension table, but Mondrian dimension. It sounded earlier in this thread like there is a Type attribute with values (Legal|Physical), and a Name attribute. Type makes sense to me as a member property of Name (i.e. Joe Smith could be of type Legal). It also makes sense as a separate cube dimension, such that you can easily filter by one or more types. Feels pretty odd to me as the lowest level of a Person hierarchy, though. It seems a little bit like having Gender as the lowest level of a Person hierarchy.