FLUID-5703: Model Transformation system must not prune out execution of all transforms whose value input is undefined

Description

The model transformation system's "fluid.model.transform.doTransform" core implementation contains the following short-circuit:

// if the function has no input, the result is considered undefined, and this is returned
if (expanded === undefined) {
return undefined;
}

This is a faulty assumption in general, especially for IoC-driven transforms. However, it is valid for simple transform functions which have been "lifted" from plain functions. We need to distinguish between functions which have received a non-empty "transformSpec" (other than "type") and those which are plain functions and defeat the behaviour for the former.

Comments

• Antranig Basman commented 2015-07-03T07:01:21.014-0400

  This is not a bug - simply a result of our own forgetfulness about the meaning of our transform grade structure. Any transform can opt out of the standard input value workflow by simply avoiding including the "fluid.transforms.standardInputTransformFunction" grade. For example, the fluid.transforms.stringTemplate transform that we were originally trying to implement can just be registered with "fluid.transforms.standardOutputTransformFunction" gradeNames.

• Tony Atkins [RtF] commented 2015-07-08T08:57:32.377-0400

  Antranig, the standardOutputTransformFunction grade also seems to return nothing unless you have a value. Can you recommend another grade?

• Antranig Basman commented 2015-07-08T13:22:51.355-0400

  Hi @@Tony Atkins [RtF] - the output works just fine, as shown in the following commit - https://github.com/amb26/infusion/commit/194850705c5dbebf56f423a4847868e2f5bdb1ea (part of https://github.com/fluid-project/infusion/pull/616 ) - perhaps you have some other problem.