In Alloy, some operators are not defined for all relations. For example, transpose ('~'), range, or domain are only defined for binary relations. However, I am implementing a Java class for the new parser that models a relation. I could throw an error when these operations are executed and the relation is not binary. However, I am not sure why i could just not execute them because there is a plausible implementation for them. For example, transpose could reverse all elements in each tuple. Domain could be the first column and range could be the tail.
This does not mean Alloy models will accept non-binary relations for these operations. This would require a language change and I am not seeking that here. It is purely on the code the actual relations used in a solution and in my case for the type class.
Are there any ideas why range, transpose, domains, and maybe others only are defined on binary relations?
If you’re talking about domain and range restriction operators, they do work on higher-arity relations AFAIK.
Restriction are first/last column true. I was thinking of the util that defines range/domain.
But transpose and the closures are only on binary relations.
In general, your proposal sounds fine to me, @peter.kriens. I don’t recall why we didn’t make transpose more general; it might have been the cost of computing it. For domain and range, I do recall some hesitation over whether the domain of a ternary relation (for example) should be a set or a binary relation (the set of two element prefixes of the relation’s tuples), but that bears more on the language design than the API perhaps.
Note that closure can’t be generalized so easily though. The arity of a join is two less than the sum of the arities. So for a binary relation, a self-join has an arity of 2+2-2 = 2. But for a ternary relation, the arity will be 3+3-2=4, so the normal definition of closure won’t work.
1 Like