Unused requirements & params == invalid model? about tck HOT 7 CLOSED

An easy fix is using the SHOULD per RFC jargon and I believe is the most pragmatic suggestion we can ask the RTF to fix.

from tck.

Hi @tarilabs - Thanks. I didn't mean that actually. I was thinking from the TCK perspective. I don't think the spec needs fixing. A requirement is after all, a requirement. Changing it to 'should" would mean mean it becomes optional ... and an "optional requirement" doesn't make too much sense.

Those requirements are part of the documentation that the model represents and (IMO) need to represent the truth. Having them optional seems to undermine that. I quite like they are mandatory.

Actually, I was more meaning we should take it into account into the TCK - and as we're not testing for (and can't test for) model validity, we should consider amending or removing the tests that violate it - and maybe getting an extra validation in or something. More like that.

And .. as I am the criminal that put in the tests that violate it, I am also happy to unwind them!

from tck.

I really believe is a bug in the spec, it should read SHOULD wrt to requirements being used in the DRGElement's expression. Otherwise, people using Conformance Level 1 without FEEL could never have a compliant DMN model since they don't specify the decision logic of Decisions/BKMs.

Thank you for bringing this up, we will refer it to the OMG RTF when opening 1.5.

from tck.

hey @tarilabs - good point about level 1 - it is a shame though. As a mandatory thing it keeps the models 'pure' and kind of enforces them, as documentation, to have only valid requirements. I kind of like that.

As a thought, if the spec is to change, maybe it should keep those SHALLs and musts but restrict them to level 3 compliance.

from tck.

Well spotted @StrayAlien !

I think the spec should be changed to use SHOULD instead of MUST. No other known PL or DSL does that, is a proven good language design pattern. There is no point to try and innovate when there is no gain. If the requirements are not used the model is still functionally correct.

from tck.

Indeed as Octavian reiterates.

Moreover, Editors and Authoring tools can always adopt "tools smarts" to suggest unused requirements, similarly to common IDE suggesting unused variables (say in Java with VSCode underlined with squiggly line)

I've marked this with for DMNv1.5 label so as soon as new RTF is opened I will raise an OMG JIRA issue, following suggestion in this phase by RTF chairs.

from tck.

Reported as https://issues.omg.org/issues/DMN15-88

from tck.