[Mondrian] RES: Course grained measure groups

Kleyson Rios kleyson.rios at saude.go.gov.br
Tue Mar 18 09:58:49 EDT 2014


I'm having the same concern dealing with Multi-granularity.

On the book you says:

There’s a step between using and not using a dimension in a measure group: a
group can use a dimension at a coarse granularity. Consider the Forecast
measure group, whose dimensionality is shown in table 5.2.

Dimension	Forecast
---------	--------
Time		(time_year_id, time_month_id) composite foreign key
Customer	customer_country_id foreign key
Product	product_id foreign key

And there are no example about physical definition for the forecast fact
table, I mean, which id from the time and product dimensions I should use to
store the rows on the forecast fact table ?

As Matt said, for the same Year/Month combination on the Time dimension I
will have many id's.

Kleyson Rios
Gestor de TI
Coordenador de BI / Banco de Dados
Gerência de Tecnologia da Informação
Secretaria de Estado da Saúde de Goiás
(62) 3201-3707

Se não pode vencer pelo talento, vença pelo esforço. (Dave Weinbaum)

-----Mensagem original-----
De: mondrian-bounces at pentaho.org [mailto:mondrian-bounces at pentaho.org] Em
nome de Matt Campbell
Enviada em: quarta-feira, 26 de fevereiro de 2014 11:05
Para: Mondrian mailing list
Assunto: [Mondrian] Course grained measure groups

I'm trying to get a handle on how course grained measure groups work in
Mondrian 4.  The example in Mondrian in Action describes the scenario
where measures in one group are at day level, while in another group
they're at month level.  The schema snippet shows a compound foreign key
link with the_year and month_of_year, with attribute specified as 'Month'.

A couple questions:
1)  The book says the value at the day level for the course MG should be
"inherited".  What's the mechanism for that?  How do we avoid incorrect
rollups?  I was having trouble seeing where in the code that's handled.
2)  If the join to the time dimension uses the_year & month_of_year,
won't measure values be inflated by the number of days?  I.e. since the
dim table will repeat the month_of_year value for each day in the month.

Multi-granularity is something I've been interested in for awhile, and
some questions about it on the forum prompted me to dig a little.

Mondrian mailing list
Mondrian at pentaho.org

More information about the Mondrian mailing list