Stream: beginners

Topic: Opaque types - Type "prefix"

Patrick Wierer (Jan 01 2025 at 17:35):

Hey,
when working with opaque types, I constantly seem to get it wrong when to "prefix" my def with the actual type, e.g. when passing it to another function that expects such a type.

Example:

A := {
    input: List U8
}

new : {} -> A
new = \{} -> @A { input: [] }

someFunc : A -> Result {} _
someFunc = \@A a ->
    a2 = anotherFunc (@A a) # only works with @A
    a3 = anotherFunc a2 # only works without @A
    Ok ({})

anotherFunc : A -> A
anotherFunc = \@A a ->
    @A a # only works with @A

Is it maybe just me not understanding when to explicitly prefix the def, or is this currently a known "issue"?
Anyway, I am very new to Roc, so my limited understanding and experience with Roc might be very well the root cause.

Ayaz Hafiz (Jan 01 2025 at 17:40):

Sorry to hear that it's been confusing when to prefix the type. If it helps, the type of @A a and a itself are different. In this case @A a is the entire opaque type A. a is the type "inside" A, so { input: List U8 } in your example. To pass { input: List U8 } to anotherFunc, it must first be wrapped in the A. But anotherFunc returns an A, so you don't need to wrap it again when calling a3 = anotherFunc a2.

Patrick Wierer (Jan 01 2025 at 17:47):

No worries, it's the normal process with everything new. Just takes time to internalize things :blush:
Thanks for the explanation, it definitely makes sense to me now.
So it seems the argument @A a from someFunc is actually unwrapped, which explains my confusion as to why I would need to wrap it again when calling anotherFunc the first time.

Sam Mohr (Jan 01 2025 at 17:50):

Yep!

Last updated: Jul 06 2025 at 12:14 UTC