Possible incompatibility issue about tck HOT 5 CLOSED

Thank you everyone for your investigation. Based on the conversation it looks like we could close this issue, is that correct please?

from tck.

Hi @tarsij - if I understand what you mean, we have quite a few tests that assert that if a decision result is a context, then its values are accessed as .. https://github.com/dmn-tck/tck/blob/master/TestCases/compliance-level-3/0007-date-time/0007-date-time.dmn#L114 is an example. The decision 'Date' returns a context, and when using it, another decision refers to Date.<fieldname>.

Hopefully I understood correctly.

from tck.

Hi @StrayAlien,

Sorry for the late reply.
I think that's not exactly what I meant.

I'll try to give an example here

------------------------------------
|       |       |     DecisionA    |
|-------|-------|------------------|
|  In1  |  In2  |        Out       |
|-------|-------|------------------|

                 ^
                 |
                 |

    ---------------------------
    |      |     DecisionB    |
    |------|------------------|
    |  In  |   Out1  |  Out2  |
    |------|---------|--------|

Lets say In1 will use the Out1 and In2 will use the Out2.
In Camunda it will be referenced as Out1 and Out2 in the input expression.
In Drools it will be referenced as DecisionB.Ou1 and DecisionB.Out2.

from tck.

Hi @tarsij

Got you. Though, same still holds. Both season and guestCount decisions return contexts - each has a decision table as an expression. The decision tables both have multiple output columns with a hit policy of unique. The 'type' of a decision table with multiple outputs and a unique hit policy is a context - so therefore accessing the result values of the decision is done using . context syntax.

As far as I can see we don't have an explicit test for this scenario, but, we do have a number of tests where the input expression of a decision table is a referring to a field in a context so uses ., and we also have tests that (directly or indirectly) assert the return type of multi-output decision tables with a unique hit policy is a context.

I guess, in summary, the way Drools does it is how the TCK test asserts stuff.

As to why the difference in implementations etc maybe Camunda can chime in? @falko

from tck.

@tarsij your observation is correct. 👍 In Camunda Platform 7, the DMN engine exposes the output of the required decision as an incorrect structure.

The behavior is fixed by the Camunda DMN-Scala extension or the new DMN engine that is used in Camunda Platform 8.

from tck.