Nearly every programming language uses dot notation, with expressions such as x.y. Yet the y here is not a value, and this makes it rather unlike other expressions. I have a simple remedy.
Atoms are having a revival. Elixir is a popular language which uses them. My proposal is that atoms are created by a dot followed by letters. So .y is an atom constant. And just as f z is a function call, the same as f(z), so x.y is a function call, the same as x(.y).
The problem is that x is not usually a function, but rather some sort of structured object with named subcomponents. The solution is for x to have an automatic conversion to the right function that will pick out the required subcomponent. This is a function that will be simplified at compile time with no run time overhead. The syntax that generates our structured object can autogenerate the required conversion declarations.
The order lattice of types described in the previous post (https://wombatlang.blogspot.com/2024/11/the-case-for-subtypes-real-world-and.html) generates automatic conversions between types which have a non-empty Intersection. The conversion always goes down to the Intersection (which might fail) then up to the destination.
Ad hoc automatic conversions can be declared, but only if the source and destination have an empty Intersection. That will be the case for the conversions we want from our structured object to the function that accepts our specific atom argument.
I hope you agree that it is nicer if everything is just procedure calls with no magic special meanings.
No comments:
Post a Comment