While browsing parallel extender forums , I came across many threads which advise the use of Row generators in situations where we have to count rows and propagate count as 0 (out of a transformer) even if there are zero rows in the input dataset.
This makes me wonder, if row generator can be used in the code that will be deployed in production environment? Even if the row generator is going to generate just one pseudo row
Shouldn't row gen be used only for development/debugging purposes?
If it's the right technique to get the job done, does it matter? DS is a toolbox, with a large variety of tools and widgets....use the best one for the job at hand...and weigh the issues surrouding it. Of course you can and should use Row Generator in "production"..... provided you have considered all of the other criteria and non-functional issues that could impact it. Might it impact performance? Maybe...but you also might have a job that doesn't process enough rows to matter, and (not claiming this, just saying it's a metric) using Row Generator might be simpler for new members of your team to maintain and understand.....which could also be an important metric.
Absolutes can get you in trouble. Its like saying "never use Server anymore"....or "always avoid Transformers....."....Fallacies too, but things I've heard from time to time.
Don't be thrown off but the artifical classification of the stage in the 'Development' branch of the repository. As Ernie notes, it has a purpose and while it certainly can be used for dev/debug specific purposes, if the stage solves a production problem and you are aware of any pros/cons, who cares how it is labelled?
Here's one absolute that won't get you into trouble - Ernie is The Man. Well, the XML Man at least. Absolutely.
-craig
"You can never have too many knives" -- Logan Nine Fingers
I always had my doubts about using the row generators in my production jobs .... now that my fears have been alleviated, I feel much more free and ready to design jobs in a much better fashion!!
