How to configure cache for Viewer Type in 3.0?

As of 3.0, the Viewer type in our schema throws the Cache data may be lost... warning in the console because it does not have an id field, And I’m just not sure the best way to address it. I’m trying to determine the ideal solution here (other than removing the viewer type entirely).

If I write a custom merge function on Viewer it would need need to be written for each field individually, which seems like a bad idea.

I tried writing a custom merge on Query but that didn’t seem to work:

    new InMemoryCache({
      typePolicies: {
        Query: {
          fields: {
            viewer: {
              merge(existing, incoming) {
                return { ...existing, ...incoming };
              },
            },
          },
        },
      },
    });

If I utilize possibleTypes and map Viewer to the types of its fields (eg User Product AccountType), it doesn’t seem to fix the issue. Do I need to remove references to the viewer in my queries to allow this to work?

The only other option I can think of is adding id to Viewer, but trying to avoid that if I can.

Any input is appreciated!

[benjamn]

Ok, since it sounds like Viewer is a singleton, you can give it a constant ID (without having to invent a real identity):

new InMemoryCache({
  typePolicies: {
    Viewer: {
      // This means the singleton Viewer's identity does not depend on any of its
      // fields (except for __typename), making the ID effectively constant.
      keyFields: [],
    },
  },
})

Alternatively, the following tweak to the merge function might make a difference, since it will do a better job of merging sub-fields that have their own merge functions:

merge(existing, incoming, { mergeObjects }) {
  return mergeObjects(existing, incoming);
}

I’ll leave this issue open in case those solutions don’t work, or you have any other questions.

[hwillson]

It doesn’t sound like there is an outstanding issue here, so closing. Thanks!