track? preserve? for variables dervied from other parameters

[oralb]

I am constructing a system. Below is the brief code;

@System test(Controller) begin
---omitted upper parts---

a: a ~ preserve(parameter)
b: b ~ preserve(parameter)

alpha(a, b, c): factor => begin
    (a + b) 
end ~ preserve? track?

--- omitted lower parts-----
end

My question is about the "alpha" in the code. The alpha is derived from a and b which are "preserve(parameter)".
I believe that alpha should also be "preserve(parameter)". But "track" is OK when the system was constructed and then performed.

I wonder which one is close to the developer's intention and Cropbox's performance.

Thanks in advance!

[tomyun]

The choice between preserve and track boils down to intended behaviors of the variable that is declared. Generally, the most of preserve variables are used as parameters, but then there are still some less common cases where we want non-parameter preserve variables, e.g., automatically derived parameter or static assignment of external value. You can then create a preserve variable with computational logic, but track also should be able to do the same work as you mentioned.

In such case, I'd recommend using track. There is a chance that its depending parameters later replaced by some dynamic variables after systems composition, leading to an unexpected logical error hard to be caught up.

Technically, preserve has a performance advantage when compared to track since the former only generates code that runs once with no further update. Practically, the actual difference should be slim as we currently have major bottlenecks in other parts of the framework.