Apollo cache WriteError prevention

[mshwery]

One issue I've seen come up multiple times is Apollo's caching mechanism throwing when a given __typename has at some point been queried with id (which Apollo uses to normalize the cache), but subsequent queries (maybe upon specific types of client-side routing/navigation) omit the id for the same type.

This causes Apollo to throw an exception, treating the query as a failure even though data came back successfully. For example:

query GetMovies {
  movies {
    id
    name
  }
}

Then, clicking something or navigating somewhere leads to this query:

query GetMovieNames {
  movies {
    name
  }
}

This would throw an error not because the data couldn't be retrieved but because the cache doesn't know how to handle it.

eslint-plugin-graphql has a unique opportunity to catch these types of issues pretty easily. Thoughts on adding a validation rule to prevent this?

The outcome would be if you have any documents that include the id field for a given __typename, you could warn/error if any other documents have omitted the id field.