Type definition for `DefinitionNode` appears to be wrong

[jacob-orbiit]

Property 'selectionSet' does not exist on type 'DefinitionNode'.
  Property 'selectionSet' does not exist on type 'SchemaDefinitionNode'.ts(2339)

I've inspected parse(op).definitions[0].selectionSet, and it does indeed exist.

I see DefinitionNode includes ExecutableDefinitionNode → OperationDefinitionNode, and OperationDefinitionNode has selectionSet, but the definition doesn't seem to follow that deep.

[JoviDeCroock]

A DefinitionNode can also be part of the SchemaDefinition part of the GraphQL lanaguage, this means that to efficiently narrow the type you would have to do.

if (definition.kind === Kind.OPERATION_DEFINITION || definition.kind === Kind.FRAGMENT_DEFINITION) {}

Doing the above ensures that you are dealing with the query-language and hence have a selection-set.

[jacob-orbiit]

Ah, that does solve the initial selectionSet issue. But then it reveals:

Property 'alias' does not exist on type 'SelectionNode'.
  Property 'alias' does not exist on type 'FragmentSpreadNode'.

Property 'name' does not exist on type 'SelectionNode'.
  Property 'name' does not exist on type 'InlineFragmentNode'.

Property 'operation' does not exist on type 'OperationDefinitionNode | FragmentDefinitionNode'.
  Property 'operation' does not exist on type 'FragmentDefinitionNode'.

There surely must be a way to do this would a bunch of narrowing because the code actually works fine (perhaps because my inputs are uniform?).

[JoviDeCroock]

I mean, these are all valid type errors and your code will fail for other permutations of the query-language. 😅

the first one is the distinction between

query { x { alias: field ...fragment } }

Where you can see whether the .kind is a FragmentSpreadNode, same goes for the second where the distinction is between an inline fragment and a selectionNode as an inline fragment has no name only a potential type-condition and the last one is the distinction between defining an operation and defining a fragment. All of these are valid issues to point out

[jacob-orbiit]

Ahh, yes. Sorry, of course 🤦‍♂️ Thanks for the explanation.