Here old FE had a better error message.

[massimo ceraolo]

Consider the following code:

model a
 Real x=1;
 parameter Integer N=2;
 parameter Real R1[N, N](unit="Ohm/m") = 0.3*[1, 0; 0, 1]/1000
          "resistances per unit length";
equation
end a;

Old FE issues the following warning:

[1] 08:35:42 Translation Warning
[a: 4:26-4:38]: Non-array modification '"Ohm/m"' for array component, possibly due to missing 'each'.

which is very clear. For the same code the new FE issues an error, with the following message:

[2] 08:36:40 Translation Error
[a: 4:2-5:40]: Type mismatch in binding unit = "Ohm/m", expected subtype of String[2, 2], got type String.

I don't know whether this should be a warning or an error. Anyway, the Old FE message is clearer from the user point-of-view. I propose a message similar to the old one, or a combination of old and new ones.

[Francesco Casella]

Replying to ceraolo:

I don't know whether this should be a warning or an error.

It is indeed an error, because the code is invalid. R1 is a two-dimensional array, so the unit modifier should also be a two-dimensional array of the same dimension, as suggested by the error message.

Alternatively, you can use 'each', as in

model a
 Real x=1;
 parameter Integer N=2;
 parameter Real R1[N, N](each unit="Ohm/m") = 0.3*[1, 0; 0, 1]/1000
          "resistances per unit length";
equation
end a;

that compiles and runs with the NF.

The old FE was a bit more tolerant on missing 'each', but that could have led to problems, so we decided to be strict with the handling of such modifiers, according to the specification.

@perost, in these cases (type error in bindings due to wrong dimensions) could we perhaps add a suggestion to add 'each' in the error message? That could help a lot understanding how to fix the problem.