[Mondrian] RE: Eigenbase perforce change 13913 for review

Julian Hyde jhyde at pentaho.com
Wed Nov 17 15:17:06 EST 2010

Please don't do that. We are always getting test failures caused by someone
having updated logs for databases X and Y but not Z.
Even if log files weren't an issue, I believe that we should define data
types in the mondrian schema, not inherit them from the database schema.
People might want a measure stored in a NUMBER(38, 0) column in Oracle to be
represented as a java int. We should let them.
In physical schemas, to be introduced in mondrian-4, you will be able to
define the data types of the columns.


From: Luc Boudreau [mailto:lucboudreau at gmail.com] 
Sent: Wednesday, November 17, 2010 11:59 AM
To: jhyde at pentaho.com
Cc: Mondrian developer mailing list
Subject: Re: Eigenbase perforce change 13913 for review

Yeah, I saw that after I checked it in. What I'll do is the same approach as
the SQL pattern detection tests. I want to extend the XML ref logs and add
an attribute to tell which response XML corresponds to which test context
database dialect. This way we won't have to bend the code in order to
support the test cases. You said that the original design was to avoid
having multiple ref logs, what about having something like:

    <response dialect="default">
    <response dialect="oracle">

On Tue, Nov 16, 2010 at 5:20 PM, Julian Hyde <jhyde at pentaho.com> wrote:

> Luc Boudreau wrote:
> http://p4web.eigenbase.org/@md=d
<http://p4web.eigenbase.org/@md=d&c=6PU@/13913?ac=10> &c=6PU@/13913?ac=10
> Change 13913 by lucboudreau at luc-mondrian-3.2 on 2010/11/16 06:48:42
>       MONDRIAN-613 Missing xsd:type attribute in drill-down
> rowset makes Excel 2007 crash. Also, when returning a
> drillthrough query result, use the actual DB specified type
> for the row set XSD infos rather than the schema infos.


The trouble with inferring the XML type from the database type is that the
database type is different for every database. So, your change has broken
the drillthrough tests in XmlaBasicTest on oracle. The previous behavior was
mostly a matter of expediency, to avoid maintaining multiple ref logs, one
for each database.

Any ideas how to fix this?

As it happens I'm working on XMLA drillthrough support right now, so the
test exceptions had me mighty puzzled for a while.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.pentaho.org/pipermail/mondrian/attachments/20101117/33fec9a7/attachment.html 

More information about the Mondrian mailing list