perrformance issues using transformer

[pgkumar]

Hi

Our best prastices make us use builops instered of transformer stage even for trimimg of spaces. Please let me know whether builops are very efficient than transformer stage in all the cases?
Can you give some senarios where transformer performs well than builtops.

Thanks in advance
Regards,
Varun

[kumar_s]

Hi Varun,
Welcome to Dsxchange :D !!!
Its not true for all the cases. Ray has mentioned in one post that, IBM has announced that performance of Transformer stage is quiet superior that a Filter stage for some cases. So it would be easy to find the results by yourself. Construct two jobs, one which uses the transformer and other uses the Buildops. And dont forget to post the result here.

[pgkumar]

Thanks Kumar

I will try it out post the same here

Cheers,
Varun

[ray.wurlod]

The transition point is version 7.1; it's in this version that IBM claim to have enhanced performance of the Transformer stage.

So, how good is your C++ compared to theirs? You can view the generated C++ if you wish; it's in a subdirectory called RT_SCnnnn, where nnnn is the job number. (RT_SC = runtime source code.)

[DSguru2B]

Performance is a relative term. A job that runs for half an hour in a 9 hour time window might be considered efficient where as a job that runs for 5 mins in a 10 min time window, might not. A build op for everything, as little a work as trimming is, IMHO, to much extra work. If that were the case I'd trim the fields in the database stage itself. If it were a flat file, I'd use a filter command command.
Transformers are good. Use 'em. Not to go against your in-house standards but if you can change them for ease of maintenance, go for it.

[ray.wurlod]

The Modify stage also has a string_trim() function. This may be faster than either the Transformer or the Build stage.