Snowflake model - modeling

[#1]

Hi Experts,

I am very much new to QlikView.  I have a following scenario.

1 FACT table , 1 Main Dimension , 3 sub dimensions

Main Dimension Vs FACT - 1:N

Main  Dimension Vs 1st Sub Dimension-  1: N

Main Dimension Vs 2nd Sub Dimension - 1:N

Main Dimension Vs 3rd Sub Dimension - 1:N

How can i join these tables in my design?  I am trying following -

FACT LEFT JOIN  Main Dimension

LEFT KEEP 1st Sub Dimension

LEFT KEEP 2nd Sub Dimension

LEFT KEEP  3rd Sub Dimension

Please advise on my design.

[#2]

Hi, You really don't need to join these table. Qlikview is a association type tool and so the table will be automatically connected if the 2 tables have same field names. Please load the tables with out synnthatic keys make sure that the keys between the join correct.

Please post sample data with field names to help further.

[#3]

Thanks Dathu. But as per my team's standards, we need to have a single QVD file with all the processing. That is the reason, i am joining main dimension and FACT. Rest of the tables, i am using LEFT KEEP.

I have ensured that the design has no Synthetic keys.

My question is  ---- Do in use RIGHT KEEP or LEFT KEEP  for joins between Main Dimension Vs Sub dimensions.

Thanks in advance.

[#4]

Hi Devendran, Basically We consider FACT table as our primary data. So follow LEFT KEEP to remove the unnecessary data from Dimension tables as dimension tables have more data.

[#5]

I appreciate your prompt response.

Yes. I am using LEFT KEEP to associate my tables. My concern is - I might lose some data which are present in sub dimensions  when i join Main Dimension (1:N) Sub Dimensions. I don't want to use RIGHT KEEP which will increase my QVD size as well may induce wrong results.

As of now, i see higher count for the FACT table key field in the final design. I am yet to test the data but wanted to ensure that i am in the right path as Modeling is the base.

Thanks in advance.

[Joseph_Musekura]

Hi,

Pease check this link: Understanding Join, Keep and Concatenate

http://community.qlik.com/message/166044#166044

Regards