Confusion about how to generate output file when there is a custom scalar type

[oveddan]

I have a schema with custom scalar type BigInt:

type TransactionInfo @entity {
  id: ID!
  chainId: BigInt!
  network: String!
  block: BigInt!
  logIndex: BigInt!
  address: Bytes!
  timestamp: BigInt!
}

When I try to run gql-tada generate output with this schema, I get an error "Error: Unknown type: "BigInt"."

How can I instruct, in the generate output step, how to handle these custom scalars? I see there is a section on custom scalars, but that seems to only apply after the output is generated?

[kitten]

Do you actually have a scalar BigInt definition in your SDL file? That error sounds more like the schema parsing has failed 😅

[oveddan]

We're currently using the Graph Protocol's tooling to generate types from the schema. It supports their custom scalar types explained here.

We use the command graph codegen and it works.

I'm not quite sure the internals of how they set it up to parse this custom scalar type. Do you have any idea of how to integrate with gql.data?

For reference, graph codegen task is able to successfully generate a schema.ts file.

The package where we are attempting this is here.

[JoviDeCroock]

Hmm, that would actually imply that the schema you have is invalid as it is missing a scalar BigInt definition. 😅

You could manually pull down the schema by doing gql-tada generate schema and then adding the scalar BigInt yourself.

That being said with a valid schema you would just have to add your scalar to your initGraphQLTada config and we would translate it to the right type. https://gql-tada.0no.co/reference/gql-tada-api#initgraphqltada

[oveddan]

Thanks for the quick reply. It seems like a bit of a chicken and egg issue - I can't run pnpm gql-tada generate schema ./schema.graphql - when I do I get the same error Error: Unknown type: "BigInt". because the scalar type is not defined in the schema. I believe the graph protocol's tooling injects that custom scalar type dynamically when doing codegen.

And correspondingly, I cannot setup initGraphQLTada because it requires the introspection object from graphql-env.d.ts which has not been generated yet.

[oveddan]

It would be great if you could specify custom scalar types at runtime when generating the output file.

[kitten]

Can you look into whether the schema that you get from the Graph is valid please?

If you have an SDL file, or if you pull one using generate schema:

• does it contain scalar BigInt as one line?
• if not, does it reference BigInt as in the example you've posted?
• does it parse with buildSchema?

There's some other notes in other discussions about this, but we don't accept incomplete SDL (i.e. no modifications or split up SDL files will be pulled in) because we're not making a schema builder, and there's too many ways to build schemas. In other words, SDL/introspection query results are basically a standard way of conveying a complete schema, so we don't want to instead also support all the ways of building a schema "from scratch" 😅

In your case it's of course presumably not built from scratch, but I do want to get to the bottom of this. If the schema is invalid, we can maybe add a workaround and automatically add scalars for missing output and input types (although, I have to think about this further, this might be dangerous and confusing)

But primarily it sounds to me like a third-party bug, so it might be better to confirm the above and report it to The Graph