Hacker News Reader

Comment by brandonbloom | Hacker News Reader

> you can't make any type error when unwrapping the enum

This is also true in Go with type switches:

    switch y := x.(type)
        case MyStruct:

And with type assertions:

    y, ok := x.(MyInterface)

In both cases, y is of the correct type. The ok is optional, if omitted and the assertion fails, you'll get a runtime panic.

View on HN

y is of the type you asked for, yes, but that isn't necessarily the type that should be there if the programmer makes a mistake, or if the underlying code changes. With an enum, the types are written in one place, and the compiler connects the dots, giving compile-time errors about mistakes/mismatched expectations. This similar to how one can call `foo(x)` and the return value's type is known, no need to first cast the function to the type the programmer thinks it should be. The casting pushes type errors to execution-time rather than compile-time.

> The casting pushes type errors to execution-time rather than compile-time

Which is exactly what pattern matching does, which is what I was responding about.

Ok, i'm starting to understand what we're disagreeing on. You're talking about pattern matching with dynamic expressions ( don't know the official name for that), such as what you find in erlang, i'm talking very basic enum unwrapping based on types, which is compile time checked. Swift also adds some kind of dynamic expression to patterns, which may break exhaustiv checking, but in the case of go, just a simple type check would be a great start.

Edit ( since i can't reply) : just look at https://developer.apple.com/library/content/documentation/Sw... with something like the barcode example.

It's all compile time checked.

> very basic enum unwrapping based on types, which is compile time checked

Sooo just accessing fields on a struct? I don't understand the distinction you are trying to make.

It's like accessing fields on a struct, where you can only access the ones that are guaranteed to be valid. For instance, in existing Go, the struct method would be having two fields, * X and * Y, where (in theory) exactly one of them is non-nil, but both can still be accessed. This requires manually maintaining this invariant (and remembering to check when accessing), as well as sacrificing performance by allocating. Using an ADT would allow storing the X and Y by value, as well as guaranteeing that the invariant of "X xor Y" is always true.

Sounds like type-assertions cover that use case just fine. However, you are correct that you can't do this without allocating. ¯\_(ツ)_/¯

As people keep saying, type assertions do not have as strong compile time assurances as ADTs.

And as I keep saying: I don't value those assurances and in fact find them to be a net loss when accounting for breaking changes changes to clients.

You personally not valuing the differences is very different to not having any differences, and the latter is what a lot of your comments seem to be implying, including the one I was replying to.

You pay the price of getting the bugs and edge cases out regardless of whether you have compile time assurances.

The complexity is inherent and you can either choose to make it implicit and play the losing game of ensuring it doesn't fail with tests or make it explicit from the outset.

I'm sorry, i meant trying to unwrap to a type that's not even possible ( and forgeting to check for one that is). Sure, there are ways to dynamically ensure typecasts have succeeded.

Go will give you errors for "impossible type assertions" in many cases.

> The ok is optional, if omitted and the assertion fails, you'll get a runtime panic.

Great, now my enums can panic because I have to play "be the typechecker", instead of focusing on something actually important like business logic.

How is this not seen as universally a bad thing?

Reply on news.ycombinator.com