(trying to adapt <https://stackoverflow.com/questi...

# announcements

[pniederw]

(trying to adapt https://stackoverflow.com/questions/36253310/how-to-get-actual-type-arguments-of-a-reified-generic-parameter-in-kotlin to get a KType rather than a Type)

[Marc Knaup]

Do you mean this one?
https://gist.github.com/udalov/bb6f398c2e643ee69586356fdd67e9b1

It somehow works for me.
The ERROR type I only get if I don’t create an anonymous subclass at the call site (i.e. in the inline fun getKType).

I noticed however that the resulting KType (and KTypeProjection) has various issues:
• You cannot get the KClass or Class which was used to create the KType so you have to store that info separately.
• The KType lacks nullability information.
• The KTypeProjection doesn’t have star projection anymore (turns into out UpperBound? at runtime).
• KType itself isn’t generic and thus you lose compile-time safety when passing around `KType`s instead of `KClass`es.
• I think `KType`’s implementation isn’t optimized that much for performance. isSubtypeOf for example looks scary.
• KTypeProjection lacks isSubtypeOf or something similar to check whether two projections are compatible/assignable.

[Marc Knaup]

In the end it all depends on your use case.
I was lucky because I only need a subset of `KType`’s information and wrote my own very specific wrapper:
https://github.com/fluidsonic/fluid-json/blob/6594ab0bfe66252a00832b612fc76bc1da24401e/Sources/JSONCodableType.kt

It basically ignores variance and nullability since they’re not reliable anyway. It pre-computes the hashCode for performance and caches ParametrizedTypes for the same reason.
It doesn’t support some Type classes and multiple upper/lower generic bounds since they’re all not relevant in my case.

[pniederw]

The toString() of the resulting KType always contains “ERROR: …” for me. I was particularly interested in the nullability information, but as you said, it’s not there.
kotlin-reflect feels a bit shaky and doesn’t seem quite there yet. also got lots of errors when debugging it.

[Marc Knaup]

Unfortunately there cannot be nullability information when KType is constructed from Java’s Type since Java doesn’t have any info about nullability.

[pniederw]

I didn’t construct KType from java’s Type, I constructed it using kotlin reflection. same thing.

[Marc Knaup]

How did you construct it? KClass.createType allows you to specify nullability 😮

[pniederw]

lol

[pniederw]

I had tried KClass.supertypes.single().arguments.single().type!!

[Marc Knaup]

Ah, yeah. Regarding generic inheritance I think Kotlin has the same limitations as Java - i.e. type erasure 😕

[pniederw]

this is using reified super type token, and works for Class. why can’t it work for KClass too?

[pniederw]

KClass.supertypes seemed like the equivalent of Class.getGenericSuperclass()

[Marc Knaup]

Yeah that would be great but I think it isn’t 😞

[Marc Knaup]

KType is mostly useless at the moment imo.

[pniederw]

I simply think it’s not mature, some things haven’t been implemented yet.