why isn't this implicitly forced?

# getting-started

[Michael de Kaste]

why isn't this implicitly forced?

[ephemient]

1. the compiler is pretty limited in what it tracks between variables for smart-casting, e.g. https://kotlinlang.org/spec/type-inference.html#bound-smart-casts
2. are you intentionally relying on invariance? if the declaration were Type<out T> then doSomethingWithType<Any>(Type.OnlyWithString(""), Holder(0)) would be completely legal

[Michael de Kaste]

Well the actual Type as is in our application is covariant (out T), but in general, I only want to be able to call doSomethingWithType if both Type and Holder have the exact same types

[ephemient]

then you can't check that they're both the same without something like data class HolderWithType(val type: String, val value: String) : Type<String>, Holder<String> and using that instead

[Michael de Kaste]

the use-case is, we're building search-criterias for our database where we have a searchCriteriaType (IsLessThan<T : Comparable<T>>, IsNull<Nothing>, etc.) and a path source from javax. For some of the criteriabuilder functions like the actual "greaterThan", the types of the path and expression need to have the same subtype. But for a like function, path needs to be string, so we setup a searchcriteria that only accepts strings as well

[Michael de Kaste]

but no matter how I implement that, it appears I can't make it "type-safe"

[ephemient]

nope

[ephemient]

it may be safe but that is only known outside of the type system