[new-parser] Improvement candidates

[tkobayas]

This issue aggregates improvement candidates. It may be something wrong or too tolerate in the old Antlr3 parser. Also it may be to drop features which would improve the maintainability. They will need to be discussed in dev at kie.apache ML.

• In the old parser, duration attribute accepts any chunk text and expect validation at the compiler phase. We may define more strict parser syntax for the duration attribute.

  • Discussion: [incubator-kie-drools-5938] [new-parser] Tests using wrong duration a… #5969
  • [new-parser] Tests using wrong duration attribute format #5937
  • [new-parser] Support ISO 8601 format for duration attribute's value #5938

• Define expected value format where we use chunk, so we can reduce ambiguity (e.g. unexpected greediness)

  • Discussion: [incubator-kie-drools-5948] [new-parser] Broken testIncompatibleListO… #5975 (comment)

• Drop Half constraint

  • e.g. name == “Mark” || == “Mario”, age > 10 && < 20
  • It creates ambiguity between “conditionalOrExpression” and “orRestriction” / “conditionalAndExpression” and “andRestriction”
  • Currently we have a fix. See [incubator-kie-drools-5879] [new-parser] build failure on kie-dmn-val… #5899
  • This syntax is not widely used

• Drop Annotation inside LHS (except @watch)

  • For example, having an annotation on AndDescr makes sense? https://github.com/apache/incubator-kie-drools/blob/main/drools-test-coverage/test-compiler-integration/src/test/java/org/drools/compiler/integrationtests/AnnotationsTest.java#L282-L298

• Drop LanguageLevelOption

  • DRL5 : Very old
  • DRL6_STRICT : Not widely used. Not in document. See [new-parser] Strict annotation parsing #5920

• || && as alternatives to infix or/and

  • Increases ambiguity. Not written in docs. See [new-parser] || and && should be allowed as alternatives to infix or and and #5936

• Additional syntax: ‘end’ has to end with a new line (may have trailing white-spaces / comment)

  • In other words, the next rule should not start in the same line (e.g. rule R1 when then end rule R2 when then end)
  • This limitation would simplify Lexer. Currently, Lexer does complex lookahead (for the backward compatibility) https://github.com/apache/incubator-kie-drools/blob/dev-new-parser/drools-drl/drools-drl-parser/src/main/java/org/drools/drl/parser/antlr4/LexerHelper.java#L59-L160
  • [new-parser] Reconsider RHS_DRL_END Lexer Island mode or parser rule #5856

• Custom Operator

  • The current parser rule {(helper.isPluggableEvaluator(false))}? id=IDENTIFIER { helper.emit($id, DroolsEditorType.KEYWORD); } is the combination of a semantic predicated and a widely matching "IDENTIFIER". It seems to have underlying risk of failure of alternative prediction in antlr4 which doesn't have syntactic predicate. We may reconsider the syntax or the implementation approach.
  • [new-parser] Adapt drlIdentifier to operator_key and neg_operator_key in DRL6Expressions.g4 #5874

---

Below are reviewing and maintainability improvements not by affecting features and backward compatibility.

• [new-parser] DRLParser.g4 and JavaParser.g4 unification #5799
• [new-parser] review drlExpression and drlPrimary in DRLParser.g4 #5807
• Implement Visitor pattern for DRL6Expressions.g4

[yurloc]

• Improvement candidate: [new-parser] unit without package should not be allowed #5935.