shpoont opened a new issue, #7238: URL: https://github.com/apache/incubator-kie/issues/7238
### Describe the bug Drools DMN 10.2.0 reports a decision-table gap for a finite numeric input domain even though every allowed value matches a rule. The reduced example has one input, `score`, with allowed values `0, 2, 4`. The two UNIQUE rules are `<= 0` and `>= 2`. All three allowed values return the expected decisions, but table analysis reports `DECISION_TABLE_GAP` with one gap. ### Expected behavior No gap for the domain `0, 2, 4`: there is no allowed input between 0 and 2. If the domain is changed to `0, 1, 2, 3, 4`, the same rules should report a real gap because the now-allowed value `1` matches neither rule. ### Actual behavior Schema/model/compilation validation completes without ERROR diagnostics. The analyzer retains a discrete domain containing exactly the declared values, with `error=false`. | Domain and rules | Native decision results | Analyzer result | | --- | --- | --- | | `0, 2, 4`; `<= 0` / `>= 2` | `0 -> none`, `2 -> full`, `4 -> full` | One gap, no overlaps | | `0, 2, 4`; equivalent literal cells `0` / `2, 4` | Same three decisions | No gaps or overlaps | | `0, 1, 2, 3, 4`; original comparison cells | `0 -> none`, `1 -> null` (no match), `2/3/4 -> full` | One gap, no overlaps; expected for this control | The literal variant is a diagnostic control, not a proposed requirement to rewrite the original table. ### How to reproduce Component: Drools DMN, `kie-dmn-validation` 10.2.0. Save this invented example as `numeric-sparse-gap.dmn`: ```xml <?xml version="1.0" encoding="UTF-8"?> <definitions xmlns="https://www.omg.org/spec/DMN/20230324/MODEL/"; id="model" name="Booking" namespace="urn:sme:dmn:fixture"> <inputData id="input_score" name="score"> <variable id="var_score" name="score" typeRef="number"/> </inputData> <decision id="decision" name="Refund"> <variable id="var_refund" name="Refund" typeRef="string"/> <informationRequirement id="needs_score"><requiredInput href="#input_score"/></informationRequirement> <decisionTable id="table" hitPolicy="UNIQUE"> <input id="column_score"> <inputExpression id="expr_score" typeRef="number"><text>score</text></inputExpression> <inputValues id="domain_score"><text>0, 2, 4</text></inputValues> </input> <output id="column_refund"> <outputValues id="domain_refund"><text>"full", "none"</text></outputValues> </output> <rule id="rule_low"> <inputEntry id="low_test"><text><= 0</text></inputEntry> <outputEntry id="low_result"><text>"none"</text></outputEntry> </rule> <rule id="rule_high"> <inputEntry id="high_test"><text>>= 2</text></inputEntry> <outputEntry id="high_result"><text>"full"</text></outputEntry> </rule> </decisionTable> </decision> </definitions> ``` The validator API invocation used for the observations is: ```java var validator = DMNValidatorFactory.newValidator(List.of()); try { var messages = validator.validate(new StringReader(xml), DMNValidator.Validation.VALIDATE_SCHEMA, DMNValidator.Validation.VALIDATE_MODEL, DMNValidator.Validation.VALIDATE_COMPILATION, DMNValidator.Validation.ANALYZE_DECISION_TABLE); messages.forEach(System.out::println); } finally { validator.dispose(); } ``` The case observations were obtained separately with `kie-dmn-core` 10.2.0, `DMNRuntimeBuilder.usingStrict()`, explicit `RuntimeModeOption("strict")`, runtime type checking enabled, and `evaluateById(..., "decision")` using numeric inputs. All inputs were within each variant's declared domain. ### Environment - Drools / `kie-dmn-core` / `kie-dmn-validation`: 10.2.0. - Java: Amazon Corretto 25.0.3. - OS: macOS, arm64. - Execution: direct Java invocation with prepared 10.2.0 JARs; no Quarkus, Spring Boot, Kogito application or GraalVM. - The unreleased main branch has not been built or executed for this report. Is this a known limitation of gap analysis for discrete numeric domains, or should the comparison form produce the same coverage result as the equivalent literal form? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
