[Mondrian] AggGen question...

Julian Hyde julianhyde at speakeasy.net
Thu Oct 25 23:29:20 EDT 2007


I'm not an expert on AggGen. If there are any experts out there, please
chime in. Your changes sound reasonable, so I've checked them in as change
10108.
 
Julian


  _____  

From: mondrian-bounces at pentaho.org [mailto:mondrian-bounces at pentaho.org] On
Behalf Of Ati Rosselet
Sent: Thursday, October 25, 2007 3:23 AM
To: Mondrian developer mailing list
Subject: Re: [Mondrian] AggGen question...


2 questions/fixes...

1.  is there a reason for using a toUpper() on the column names in
aggregates - AggGen.java.  At least in postgres, it makes the sql
invalid...(I've removed them at the moment and all seems fine - see below..
no errors raised in testsuite) 

2.  regarding previous problem.  As I see it, each AggGen is created on the
fly, so if we let the AggGen know what cube it is dealing with, we can
ignore measures in the Table from other cubes.

i.e.
FastBatchingCellReader.java 
614: final AggGen aggGen = new AggGen(cube.getName(),cube.getStar(),
columns);

and AggGen.java:55
// name of cube concerned - ignore measures from other cubes
    private String cubeName;
    
    public AggGen(String cubeName, RolapStar star, RolapStar.Column[]
columns) {
        this.cubeName=cubeName;
    ...

in init():301
                if (rMeasure.getCubeName()!=cubeName)
                    continue;
 
this seems to fix the AggGen problem..  at least this one.. and running the
test suite with these changes produces NO errors.. (except for 1
postgres/grouping sets related.. which was there before) 

also... if the Promotion sales measure is included (uses measure
expression), AggGen fails since it cannot find the column specified by the
expression... any ideas?
maybe a baseColumn property for Measure expression or something? (just
throwing ideas here). 


Haven't got around to the test creation yet...   a bit confused..

Cheers
Ati



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.pentaho.org/pipermail/mondrian/attachments/20071025/8632c748/attachment.html 


More information about the Mondrian mailing list