Incorrect definition of statement within the Modelica grammar

[mdempse1]

I have recently been implementing a Modelica parser and ran into an issue with the grammar as a result of the definition of statement within the grammar. The rule is defined as:

statement :
  ( component-reference ( ":=" expression | function-call-args )
    | "(" output-expression-list ")" ":="
      component-reference function-call-args
    | break
    | return
    | if-statement
    | for-statement
    | while-statement
    | when-statement
  )
  description

However, this then causes parsing to fail on statements such as:
der(a) := expression;

This is because component-reference does not allow der(a) on the left hand side of the assignment.

The problem can be fixed by replacing component-reference with result-reference in the statement definition. result-reference is defined within the grammar but does not appear in any of the rules and hence it is never used within the grammar.

[HansOlsson]

One might also view that as an extension of the grammar, but I agree that it makes sense to allow some variant of this.

I was a bit curious regarding result-reference - and found that it is used for the figure-annotations in the text, but, as you note, currently not in the grammar.

However, it also allows higher order derivatives (the integer part) - and I don't think that's such a good idea.

[henrikt-ma]

This is a duplicate of #2984.

[mdempse1]

Agreed, allowing assignment to a higher order derivative does seem like a bad idea to allow.

As @henrikt-ma has noted, this issue was discussed a few years ago. Claytex does have a case where this syntax is used but because Dymola accepts statements like this we didn't notice until this current project where we have a need to parse Modelica outside of Dymola.

[henrikt-ma]

Agreed, allowing assignment to a higher order derivative does seem like a bad idea to allow.

One needs to distinguish support for der(x, 2) := 1 and support for der(der(x)) := 1.

We would also need to consider der's cousin pre. It would have some application in initial algorithms to assign to pre(x), so if this is not allowed I think we at least need to include a non-normative explanation of why.

[HansOlsson]

Language group: Seems ok to add der(x).