Mappings Vs Mapplets

[naveenshindhe]

Hi,
Can anybody please let me know
1. difference between mapping and mapplets.
2. Where in project u should go for Snowflake and Starshema.
Thanks
Naveen

 

[Dagon]

1) A mapplet is a re-usable group of transformations which you can call from one or more mappings. It can't be run independently and it can't have a source or target. Instead, you pass in a set of input parameters from the parent mapping(s) and pass back a set of output parameters.

2) This is largely a matter of opinion and different people have different views on it. Star schemas will be more efficient because you are cutting down the number of table joins by not normalizing the dimension tables. However, it could mean a lot of redundancy in the data with columns being repeated and greater scope for inconsistencies in the data.

For example, if your product table has product_type and product_type_description in it, there is nothing to stop one product having product_type 'A' and type_description 'B' but another having product_type 'A' and type_description 'C'. If you normalize product_type into a separate table, this can't happen.

My rule is that if your dimension tables are simply copies of data that is on external systems, then there is no point in having them normalized. However, if they are maintained within your system, normalization is better so that you can have control the quality of the data and build the maintenance screens properly.