Possibly Incorrect Generated Type for `disconnect` in 0.27.1 about typegraphql-prisma

[MichalLytek]

@geisterfurz007
The @oneOf does not make an input union.
In this case, we can't have disconnect field of type Boolean | SomePrismaType.
We would need to split them into disconnectBoolean: Boolean and disconnectComplex: SomePrismaType and then in resolver merge them into disconnect Prisma Client query object.

[MichalLytek]

This is a big problem with graphql not supporting input union while prisma does. So they accept boolen or filter object while in graphql we can't, so in most cases it accepts the object but for this case it might be problematic.

[andrewleverette]

Thank you for the quick response. So this is a limitation of GraphQL? Do you know of a good workaround? Or is there any work that could be done to resolve this issue? I would be open to contributing, but not sure where I would start with this.

Also, just for reference. I found this document that seemed relevant -> https://github.com/graphql/graphql-wg/blob/21e8ecd30934843cbb175758eaa3d80325cd0929/rfcs/InputUnion.md

[geisterfurz007]

There is also the oneOf directive on inputs which is still not part of the official spec but supported by most tooling afaik. Maybe that could work here? I am not sure if that is an option since it would mean an object with two fields instead of a single field passed directly as argument.

[andrewleverette]

Is it possible to distinguish between relational types like one-one, one-many, etc? It seems like in the cases of a one-one relationship, the disconnect field could be the boolean and a where input type in the other cases. I could be missing something though.

[andrewleverette]

For further information, I upgraded to 0.27.0 and this issue does not exist in that version.

[shawnjones253]

For further information, I upgraded to 0.27.0 and this issue does not exist in that version.

double-checking, you don't have this problem with 0.27.0?

we do have this problem with:

"typegraphql-prisma": "0.27.0",
"prisma": "4.16.2"

[andrewleverette]

@shawnjones253, I confirmed we have the same problem in 0.27.0. There was an issue in my build process that was pulling an older version, so we were getting a false positive. We ended up just reverting 0.25.1.