Crosscut

I wonder if my idea from yesterday, to always automatically apply any function A -> B to any variant { A, ... }, turning it into a variant { B, ... }, is practical. If it is, that has profound consequences for error handling, and potentially other areas of the language.

A few months back, I published a series about algebraic effects. While I found a lot to like about them, it turned out they don't interact well with linear types. So I've been thinking about alternatives.

Using variant types in this way could be a solution, at least for error handling. We wouldn't need an error effect, or early returns, nor would we need to deal with monads. Functions could be straight-forward, linear series of instructions, with a single exit point at the end.