Should _all_ GraphQL scalars be present in a schema?

[rmosolgo]

Let's say I'm working on my BCaaS app (bean counter as a service). My only inputs are Ints and my only return values are Ints. Should my schema still contain GraphQL scalars like String and Float? Or is it appropriate to leave those types out, since they aren't used?

It looks like graphql-js leaves out unused scalars. This approach makes sense to me. GraphQL's predefined scalars aren't "universals", instead, they're saving the user a step of defining their own. In that way, like an unused custom scalar, if a predefined scalar isn't used, it's not present in the schema.

Is that intended? Or should all GraphQL schemas contain the predefined scalar types?

[stubailo]

I wonder if GraphQL lets you override the default scalars. I would guess "no" since they are used in introspection, which should be consistent across all servers?

[leebyron]

A Schema only needs to include the scalars that are present in a schema. However note that introspection is part of the schema, and the introspection system uses String and Boolean, so those are essentially always going to appear in every schema.

[leebyron]

And @stubailo, by the spec, as long as you have a type named String (as an example) - it must conform to the spec's description of how String works. So that doesn't leave much room for overriding.