const overload accepting a closure for memoized lazy evaluation

[robrix]

I think const might be more useful if it were marked as @autoclosure(escaping). As it is, you can’t use it for any kind of effectful case analysis because effects occur regardless of the branch taken, which defeats the purpose of using case analysis for effects.

[robrix]

This would be very much in the spirit of Haskell, since Haskell is lazily-evaluated by default.

[jspahrsummers]

The semantics of @autoclosure(escaping) are non-obvious to most folks, and it doesn't help that you'd have to read the documentation to realize that it automatically closes over your value.

I think escaping autoclosures are an anti-feature, and their use should be discouraged. If you want to support lazy evaluation, accept an explicit closure.

[robrix]

That’d look like so:

fooBarQuux.analysis(
    ifFoo: const(0),
    ifBar: const { ++x },
    ifQuux: { $0 + $1 })

Yeah, I agree—that’s much less problematic.

[robrix]

It’s hardly a constant unless it memoizes as well. I think that lines up nicely with this.