I love the HTTPCaller, SQLExecutor, and FeatureReader, but I have to admit I was confused how to make them work. It turned out I had to use a Creator transformer, which is not obvious. I'd prefer if they just worked without adding a Creator transformer.
Don't require Creator for transformers that read data
FME Form
Enter your username or e-mail address. We'll send you an e-mail with instructions to reset your password.
Agree with @redgeographics: all these transformers can be triggered by a "dummy" feature (Creator) or an actual feature read or created by the workspace. That is part of the power of these transformers.
Don't take that power away!
So I asked myself this question today, so here's my thoughts:
I agree with the fact that we should not take the power away that makes these transformers so useful. However, I think there would still be an upside to be able to use these transformers without the need to add a Creator.
What if there would be an option like a checkbox 'Initiate on time', to be able to run the transformer one time without the need to connect a Creator.
I see the fact that you need to add a Creator right now more as an non-chique way to work around the fact that you need to run this transformer one time.
I vote up!
Definitely agree we need to keep the Initiator port - makes these transformers so much more powerful. My workflows would be far more complicated without it. However when just needing to run once at start of translation, this idea makes sense. Although not for the SQLExecutor.
There is already the SQLCreator transformer for that, that runs once are the translation start
One of the benefits of having a creator attached to such a transformer is the ability to define the order that the creators run in a workbench. A typical case is where you have several "lookup datasets" with a few dozen records and then the main data set with thousands of records. By setting the creator run order it enables the lookup datasets to be loaded first then the main dataset starts processing. otherwise you may get the situation where all the main data has to load in memory before the lookup datasets are loaded and referenced.
All the votes have been transferred into this idea.