General subscripting

[HansOlsson]

Add general subscripting also to Modelica.
Closes #2659

As noted in that ticket the previously discussed blocking cases are basically irrelevant.
An alternative to stating that all operations generate arrays indexed by 1...n is to state that general subscripting transforms the arrays to such a form - the downside is that (a)[1] isn't the same as a[1] - the upside is that (a)[1] is valid regardless of how the array is subscripted.
The end effect should be similar.

[olivleno]

Thanks @HansOlsson for your proposal to resolve this issue.
@sjoelund @henrikt-ma @gkurzbach I'd like to encourage you to comment and review.

[henrikt-ma]

My main concern is the characterization of how a matrix-valued expression should be indexed.

[HansOlsson]

Phone meeting:
Should clean up types of array expressions - indexing should match the type.
Hans: As long as there is only indexing with integers it is safe to subscript with integers.
(and delay the rest).

Shouldn't close before we have new issue for type of array expressions, but it can be closed later.

[HansOlsson]

Added #3410 - note that it is even more restrictive than I thought.

[HansOlsson]

Looking more closely at the text I see that we already have very drastic restrictions on non-Integer subscripts. So, the text that looks like a new restriction here was in fact implicitly present (if the operations generated anything else it could barely be used) - I still think it good to have as a normative text, instead of changing it to non-normative.

The new issue is also opened.

[henrikt-ma]

Requesting that we avoid making statements that are dependent on the resolution of #3410.

[HansOlsson]

Requesting that we avoid making statements that are dependent on the resolution of #3410.

Should be resolved

[henrikt-ma]

Looks good. Assuming that the example listing has been verified to fit the page width.