Transformer Stage

[saikir]

Hi,

I believe the Transformer stages are haevy compared to other stages like Copy or Modify etc. Has this changed in IBM InfoSphere DataStage?

In the "Parallel Job Advanced User Guid(7.5.1)", it is written that we shoud use transformer stages sparingly and use other stages such as filter stage, Switch Stage and modify Stage.

However, in the IBM Redbook ""IBM InfoSphere DataStage Data Flow and Job Design", it is written "Use the parallel Transformer stage (not the BASIC Transformer) instead of
the Filter or Switch stages"

Am i missing the context over here or In 8 Transformer stages are lighter then other versions whne compared to switch or filter or copy?

Can any of you please explain as they appear to be contradicting each other?

Regards,
Sai

[sajidkp]

Parellel transformer will take more time to compile as it has to complie the C++ codes.

[ray.wurlod]

Things changed between the two releases. The Transformer stage now compiles to an (efficient) transform operator. This was not the case in earlier versions.

[saikir]

Hi Ray,

Thanks a lot for your answer. Does this mean that as compared to 7.x, in 8 i can use as many transformer stages as possible without affecting the job performance as in 8 the transformer stage compiles to an efficient Transform operator?

Including simple transformation like dropping columns, renaming columns etc?

Regards,
Sai

[ray.wurlod]

No. The advice relates specifically to Switch and Filter stages. The Copy stage is more efficient for renaming and dropping columns than is the Transformer stage.

[saikir]

Hi Ray,

I have one more question. May be i am confused. In 8 i assume that the Transformer stage still generates a C++ code and in addition uses an efficient transform operator as compared to 7.x

Regards,
Sai

[ray.wurlod]

Your assumption is correct. You can see this in the job score.