Write an event for a decision evaluation

[saig0]

Description

We should write an event when a decision is evaluated. The event should include all data about the decision evaluation (e.g. the decision result, the result of required decisions, etc.). The event can be consumed to analyze and visualize the evaluation of the decision, for example, by Operate and Optimize.

• create a new record DecisionEvaluationRecord
• create a new intent evaluated
• it should include all data about the evaluation that is provided from the DMN module
• it should include context data about the process instance that the decision is called from

The record could look like this:

{
  "decisionKey": 1L,
  "decisionId": "x",
  "version": 1L,
  "decisionRequirementsKey": 1L,
  "decisionRequirementsId": "x",
  "processDefinitionKey": 1L,
  "processInstanceKey": 1L,
  "elementInstanceKey": 1L,
  "elementId": "x",
  "decisionOutput": "x" (JSON),
  "evaluatedDecisions": [
    {
      "decisionId": "x",
      "decisionName": "x",
      "decisionType": "x",
      "decisionOutput": "x" (JSON),  
      "evaluatedInputs": [ (for decision tables) 
      	{
      	  "inputId": "x",
      	  "inputName": "x",
      	  "inputValue": "x" (JSON) 
      	}
      ],
      "matchedRules": [ (for decision tables)
      	{
      	  "ruleId": "x",
      	  "ruleIndex:": x,
      	  "evaluatedOutputs": [
  	    {
  	      "outputId": "x",
  	      "outputName": "x",
  	      "outputValue": "x" (JSON)
  	    } 
      	  ] 
      	}
      ]
    }
  ]  
}

blocked by #8080
blocked by #8112

[menski]

@saig0 for better understanding, how will be the result value be structured, i.e. if the hit policy returns multiple outputs, or the hit policy does aggregation?

/cc @nazlikaya

[saig0]

how will be the result value be structured, i.e. if the hit policy returns multiple outputs, or the hit policy does aggregation?

The result in decisionResult contains the result of the decision. The structure of the result depends on the type of the decision (table), or the hit policy of the decision table. (see also the discussion in Slack)

• If a decision table has a collect hit policy like sum then it has the collected result.
• If the decision table has a different hit policy then
  • it could be a primitive value (i.e. string, number, etc. - if there is one output and only one rule can match),
  • or a list (i.e. if multiple rules can match),
  • or a context/document (i.e. if there are multiple outputs).
• If the decision table is a literal expression then it is the result of the expression.

[saig0]

@sdorokhova I updated the event and added the ruleIndex (int) to the matchedRules property. The property will contain the index of the matched rule, starting from 1. It's an addition to the ruleId from the XML. It should help to avoid parsing the DMN XML to get the rule index.

[saig0]

EDIT

• added new property decisionOutput that contains the output (i.e. the result) of the called decision
• renamed property decisionResults to evaluatedDecisions
• renamed property result of decisionResults to decisionOutput